├── model3 ├── __init__.py ├── __init__.pyc ├── cls_hrnet.pyc ├── default.pyc ├── model2_graph4_hrnet.pyc ├── model2_graph4_hrnet_agcm.pyc ├── model2_graph4_hrnet_sal.pyc ├── __pycache__ │ ├── __init__.cpython-35.pyc │ ├── __init__.cpython-37.pyc │ ├── default.cpython-35.pyc │ ├── default.cpython-37.pyc │ ├── cls_hrnet.cpython-35.pyc │ ├── cls_hrnet.cpython-37.pyc │ ├── model2_graph4_hrnet_sal.cpython-35.pyc │ ├── model2_graph4_hrnet_sal.cpython-37.pyc │ ├── model2_graph4_hrnet_agcm.cpython-35.pyc │ └── model2_graph4_hrnet_agcm.cpython-37.pyc ├── default.py ├── model2_graph4_hrnet_sal.py ├── model2_graph4_hrnet.py ├── model2_graph4_hrnet_agcm.py └── cls_hrnet.py ├── maps └── pipeline.png ├── evalcodes-eccv18 ├── EvalSaliency │ ├── SMeasure │ │ ├── EvalStructureMeasure.m │ │ ├── StructureMeasure.m │ │ ├── S_object.m │ │ ├── demo.m │ │ └── S_region.m │ ├── EvalAllResult.m │ ├── EvalROC.m │ ├── Fmeasure.m │ ├── QXL_ROC.m │ └── EvalResultClassResult.m └── evalmine.m ├── utils.py ├── README.md ├── loss.py ├── misc.py ├── data2.py ├── test.py └── log_file └── graph4_decoder_hrnet_agcm_stage3.txt /model3/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /maps/pipeline.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/maps/pipeline.png -------------------------------------------------------------------------------- /model3/__init__.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__init__.pyc -------------------------------------------------------------------------------- /model3/cls_hrnet.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/cls_hrnet.pyc -------------------------------------------------------------------------------- /model3/default.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/default.pyc -------------------------------------------------------------------------------- /model3/model2_graph4_hrnet.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/model2_graph4_hrnet.pyc -------------------------------------------------------------------------------- /model3/model2_graph4_hrnet_agcm.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/model2_graph4_hrnet_agcm.pyc -------------------------------------------------------------------------------- /model3/model2_graph4_hrnet_sal.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/model2_graph4_hrnet_sal.pyc -------------------------------------------------------------------------------- /model3/__pycache__/__init__.cpython-35.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/__init__.cpython-35.pyc -------------------------------------------------------------------------------- /model3/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /model3/__pycache__/default.cpython-35.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/default.cpython-35.pyc -------------------------------------------------------------------------------- /model3/__pycache__/default.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/default.cpython-37.pyc -------------------------------------------------------------------------------- /model3/__pycache__/cls_hrnet.cpython-35.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/cls_hrnet.cpython-35.pyc -------------------------------------------------------------------------------- /model3/__pycache__/cls_hrnet.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/cls_hrnet.cpython-37.pyc -------------------------------------------------------------------------------- /model3/__pycache__/model2_graph4_hrnet_sal.cpython-35.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/model2_graph4_hrnet_sal.cpython-35.pyc -------------------------------------------------------------------------------- /model3/__pycache__/model2_graph4_hrnet_sal.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/model2_graph4_hrnet_sal.cpython-37.pyc -------------------------------------------------------------------------------- /model3/__pycache__/model2_graph4_hrnet_agcm.cpython-35.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/model2_graph4_hrnet_agcm.cpython-35.pyc -------------------------------------------------------------------------------- /model3/__pycache__/model2_graph4_hrnet_agcm.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ltp1995/GCAGC-CVPR2020/HEAD/model3/__pycache__/model2_graph4_hrnet_agcm.cpython-37.pyc -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/SMeasure/EvalStructureMeasure.m: -------------------------------------------------------------------------------- 1 | function Score = EvalStructureMeasure(Prediction, GT) 2 | Prediction=imresize(Prediction,size(GT)); 3 | d_Prediction = double(Prediction); 4 | if (max(d_Prediction(:))==255) 5 | d_Prediction = d_Prediction./255; 6 | end 7 | d_Prediction = reshape(mapminmax(d_Prediction(:)',0,1),size(d_Prediction)); 8 | 9 | Score = StructureMeasure(d_Prediction, logical(GT(:,:,1))); 10 | 11 | end 12 | 13 | -------------------------------------------------------------------------------- /utils.py: -------------------------------------------------------------------------------- 1 | def clip_gradient(optimizer, grad_clip): 2 | for group in optimizer.param_groups: 3 | for param in group['params']: 4 | if param.grad is not None: 5 | param.grad.data.clamp_(-grad_clip, grad_clip) 6 | 7 | 8 | def adjust_lr(optimizer, init_lr, epoch, decay_rate=0.1, decay_epoch=30): 9 | decay = decay_rate ** (epoch // decay_epoch) 10 | for param_group in optimizer.param_groups: 11 | param_group['lr'] *= decay 12 | -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/EvalAllResult.m: -------------------------------------------------------------------------------- 1 | function [MAE2, ClassAP, ClassAUC, ClassFScore, AUC, AP, F_score, T, P, F, ClassSMeasure, SMeasure, ClassWFmeasure, WFmeasure] = EvalAllResult(ResultDir, GTPath, GTMaskExt) 2 | [MAE, ClassSMeasure, ClassAP, ClassAUC, ClassFScore, tpr, fpr, pre, F_tpr, F_pre, ClassWFmeasure] = EvalResultClassResult(ResultDir, GTPath, GTMaskExt); 3 | T=mean(tpr,1); 4 | F=mean(fpr,1); 5 | P=mean(pre,1); 6 | MAE2 =mean(MAE, 1); 7 | AUC = -trapz(F, T); 8 | AP = -trapz(T, P); 9 | F_score = mean(F_tpr)*mean(F_pre)/(mean(F_tpr)+0.3*mean(F_pre))*1.3; 10 | SMeasure = mean(ClassSMeasure); 11 | WFmeasure = mean(ClassWFmeasure); 12 | end -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/EvalROC.m: -------------------------------------------------------------------------------- 1 | function [Precision, TPR, FPR] = EvalROC( image, hsegmap, NT ) 2 | img=mat2gray(image); 3 | 4 | hsegmap=logical(hsegmap(:,:,1));% 5 | img=mat2gray(imresize(img,size(hsegmap))); 6 | img=(img*(NT-1)); 7 | 8 | 9 | targetHist = histc(img(hsegmap), 0:NT); 10 | nontargetHist = histc(img(~hsegmap), 0:NT); 11 | targetHist = flipud(targetHist); 12 | nontargetHist = flipud(nontargetHist); 13 | targetHist = cumsum( targetHist ); 14 | nontargetHist = cumsum( nontargetHist ); 15 | Precision = flipud(targetHist ./ (targetHist + nontargetHist + eps)); 16 | Precision(end) = 1; 17 | Precision = [0, Precision']; 18 | TPR = flipud(targetHist / sum(hsegmap(:))); % true positive 19 | TPR(end) = 0; 20 | TPR = [1, TPR']; 21 | FPR = flipud(nontargetHist / sum(hsegmap(:) == 0)); 22 | FPR(end) = 0; 23 | FPR = [1, FPR']; 24 | 25 | 26 | end 27 | -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/Fmeasure.m: -------------------------------------------------------------------------------- 1 | function [F, TPR, Precision] = Fmeasure( image, hsegmap ) 2 | 3 | img=mat2gray(imresize(image,[size(hsegmap,1),size(hsegmap,2)])); 4 | 5 | hsegmap=mat2gray(hsegmap); 6 | hsegmap=hsegmap(:,:,1); 7 | positiveset = hsegmap; 8 | negativeset = ~hsegmap ; 9 | P=sum(positiveset(:)); 10 | N=sum(negativeset(:));% 11 | 12 | 13 | T=mean(img(:))+std(img(:),0); %%%%%%%%%%%%%%%%%zhu 14 | %T= 2*mean(img(:)); 15 | T=min(T,0.8); 16 | positivesamples = img >= T; 17 | 18 | 19 | TPmat=positiveset.*positivesamples; 20 | FPmat=negativeset.*positivesamples; 21 | 22 | PS=sum(positivesamples(:)); 23 | 24 | TP=sum(TPmat(:)); 25 | FP=sum(FPmat(:)); 26 | 27 | TPR=TP/P; 28 | FPR=FP/N; 29 | Precision=TP/PS; 30 | if PS==0 31 | F=0; 32 | Precision=0; 33 | TPR=0; 34 | elseif TPR==0 35 | F=0; 36 | else 37 | F=TPR*Precision/(TPR+0.3*Precision)*1.3; 38 | end 39 | 40 | end 41 | -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/QXL_ROC.m: -------------------------------------------------------------------------------- 1 | function [Precision,TPR, FPR, AUC,AP,F] = QXL_ROC( image, hsegmap, NT ) 2 | 3 | img=mat2gray(image); 4 | 5 | hsegmap=mat2gray(hsegmap);% 6 | hsegmap=hsegmap(:,:,1); 7 | img=mat2gray(imresize(img,size(hsegmap))); 8 | img=(img*(NT-1)); 9 | 10 | positiveset = hsegmap; % 11 | negativeset = ~hsegmap ;% 12 | P=sum(positiveset(:));% 13 | N=sum(negativeset(:));% 14 | 15 | TPR=zeros(1,NT); 16 | FPR=zeros(1,NT); 17 | 18 | Precision=zeros(1,NT); 19 | F=zeros(1,NT); 20 | 21 | TPR(1)=1; 22 | FPR(1)=1; 23 | TPR(NT+2)=0; 24 | FPR(NT+2)=0; 25 | Precision(1)=0; 26 | Precision(NT+2)=1; 27 | 28 | 29 | for i=1:NT+1 30 | 31 | T=i-1; 32 | 33 | positivesamples = img >= T; 34 | 35 | 36 | 37 | TPmat=positiveset.*positivesamples; 38 | FPmat=negativeset.*positivesamples; 39 | 40 | PS=sum(positivesamples(:)); 41 | if PS~=0 42 | 43 | TP=sum(TPmat(:)); 44 | FP=sum(FPmat(:)); 45 | 46 | TPR(i+1)=TP/P; 47 | FPR(i+1)=FP/N; 48 | 49 | Precision(i+1)=TP/PS; 50 | F(i+1)=TP*Precision/(TP+0.3*Precision)*1.3; 51 | end 52 | end 53 | 54 | 55 | AUC = -trapz(FPR, TPR); 56 | AP = -trapz(TPR, Precision); 57 | 58 | F=mean(F(2:end-1)); 59 | end 60 | -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/SMeasure/StructureMeasure.m: -------------------------------------------------------------------------------- 1 | function Q = StructureMeasure(prediction,GT) 2 | % StructureMeasure computes the similarity between the foreground map and 3 | % ground truth(as proposed in "Structure-measure: A new way to evaluate 4 | % foreground maps" [Deng-Ping Fan et. al - ICCV 2017]) 5 | % Usage: 6 | % Q = StructureMeasure(prediction,GT) 7 | % Input: 8 | % prediction - Binary/Non binary foreground map with values in the range 9 | % [0 1]. Type: double. 10 | % GT - Binary ground truth. Type: logical. 11 | % Output: 12 | % Q - The computed similarity score 13 | %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 14 | 15 | % Check input 16 | if (~isa(prediction,'double')) 17 | error('The prediction should be double type...'); 18 | end 19 | if ((max(prediction(:))>1) || min(prediction(:))<0) 20 | error('The prediction should be in the range of [0 1]...'); 21 | end 22 | if (~islogical(GT)) 23 | error('GT should be logical type...'); 24 | end 25 | 26 | y = mean2(GT); 27 | 28 | if (y==0)% if the GT is completely black 29 | x = mean2(prediction); 30 | Q = 1.0 - x; %only calculate the area of intersection 31 | elseif(y==1)%if the GT is completely white 32 | x = mean2(prediction); 33 | Q = x; %only calcualte the area of intersection 34 | else 35 | alpha = 0.5; 36 | Q = alpha*S_object(prediction,GT)+(1-alpha)*S_region(prediction,GT); 37 | if (Q<0) 38 | Q=0; 39 | end 40 | end 41 | 42 | end 43 | -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/SMeasure/S_object.m: -------------------------------------------------------------------------------- 1 | function Q = S_object(prediction,GT) 2 | % S_object Computes the object similarity between foreground maps and ground 3 | % truth(as proposed in "Structure-measure:A new way to evaluate foreground 4 | % maps" [Deng-Ping Fan et. al - ICCV 2017]) 5 | % Usage: 6 | % Q = S_object(prediction,GT) 7 | % Input: 8 | % prediction - Binary/Non binary foreground map with values in the range 9 | % [0 1]. Type: double. 10 | % GT - Binary ground truth. Type: logical. 11 | % Output: 12 | % Q - The object similarity score 13 | %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 14 | 15 | % compute the similarity of the foreground in the object level 16 | prediction_fg = prediction; 17 | prediction_fg(~GT)=0; 18 | O_FG = Object(prediction_fg,GT); 19 | 20 | % compute the similarity of the background 21 | prediction_bg = 1.0 - prediction; 22 | prediction_bg(GT) = 0; 23 | O_BG = Object(prediction_bg,~GT); 24 | 25 | % combine the foreground measure and background measure together 26 | u = mean2(GT); 27 | Q = u * O_FG + (1 - u) * O_BG; 28 | 29 | end 30 | 31 | function score = Object(prediction,GT) 32 | 33 | % check the input 34 | if isempty(prediction) 35 | score = 0; 36 | return; 37 | end 38 | if isinteger(prediction) 39 | prediction = double(prediction); 40 | end 41 | if (~isa( prediction, 'double' )) 42 | error('prediction should be of type: double'); 43 | end 44 | if ((max(prediction(:))>1) || min(prediction(:))<0) 45 | error('prediction should be in the range of [0 1]'); 46 | end 47 | if(~islogical(GT)) 48 | error('GT should be of type: logical'); 49 | end 50 | 51 | % compute the mean of the foreground or background in prediction 52 | x = mean2(prediction(GT)); 53 | 54 | % compute the standard deviations of the foreground or background in prediction 55 | sigma_x = std(prediction(GT)); 56 | 57 | score = 2.0 * x./(x^2 + 1.0 + sigma_x + eps); 58 | end -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/SMeasure/demo.m: -------------------------------------------------------------------------------- 1 | close all; clear; clc; 2 | 3 | % set the RoodDir according to your own environment 4 | RootDir = pwd; 5 | 6 | % set the ground truth path and the foreground map path 7 | gtPath = fullfile(RootDir,'demo','GT'); 8 | fgPath = fullfile(RootDir,'demo','FG'); 9 | 10 | % set the result path 11 | resPath = fullfile(RootDir,'demo','Result'); 12 | if ~exist(resPath,'dir') 13 | mkdir(resPath); 14 | end 15 | 16 | % set the foreground map methods 17 | MethodNames = {'MDF','mc','DISC','rfcn','DCL','dhsnet'}; 18 | 19 | % load the gtFiles 20 | gtFiles = dir(fullfile(gtPath,'*.png')); 21 | 22 | % for each gtFiles 23 | for i = 1:length(gtFiles) 24 | fprintf('Processing %d/%d...\n',i,length(gtFiles)); 25 | 26 | % load the gt file 27 | [GT,map] = imread(fullfile(gtPath,gtFiles(i).name)); 28 | if numel(size(GT))>2 29 | GT = rgb2gray(GT); 30 | end 31 | GT = logical(GT); 32 | 33 | % in some dataset(ECSSD) some ground truth is reverse when map is not none 34 | if ~isempty(map) && (map(1)>map(2)) 35 | GT = ~GT; 36 | end 37 | 38 | % for each saliency method 39 | for j = 1 : length(MethodNames) 40 | % load the saliency map file 41 | predname = [gtFiles(i).name(1:end-4) '_' MethodNames{j} '.png']; 42 | prediction = imread(fullfile(fgPath,predname)); 43 | if numel(size(prediction))>2 44 | prediction = rgb2gray(prediction); 45 | end 46 | 47 | % Normalize the prediction. 48 | d_prediction = double(prediction); 49 | if (max(max(d_prediction))==255) 50 | d_prediction = d_prediction./255; 51 | end 52 | d_prediction = reshape(mapminmax(d_prediction(:)',0,1),size(d_prediction)); 53 | 54 | % evaluate the predicted map against the GT 55 | score = StructureMeasure(d_prediction,GT); 56 | score = roundn(score,-4); 57 | 58 | % save the result 59 | resName = sprintf([gtFiles(i).name(1:end-4) '_%.4f_' MethodNames{j} '.png'],score); 60 | imwrite(prediction,fullfile(resPath,resName)); 61 | end 62 | end 63 | 64 | fprintf('The results are saved in %s\n',resPath); 65 | 66 | 67 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Adaptive Graph Convolutional Network with Attention Graph Clustering for Co-saliency Detection (GCAGC-CVPR2020) 2 | ## Pipeline 3 | ![pipeline](https://github.com/ltp1995/GCAGC-CVPR2020/blob/master/maps/pipeline.png) 4 | ## Testing code 5 | * python test.py 6 | ## Pretrained models (HRNET version) 7 | * [Baidu Cloud](https://pan.baidu.com/s/1C4pX_akexLHe7fTWtWHxlw) Fetchcode: isrw && [Google Cloud](https://drive.google.com/file/d/1LvSTdWyrWhO4ge1-Cj5ETE2SrhvebIiR/view?usp=sharing) 8 | ## Training Dataset (COCO-SEG, 78 categories, 200K images) && Cosal results 9 | * [Baidu Cloud](https://pan.baidu.com/s/1_8wDB52k-FUqUZHTMCZ2vA) Fetchcode: rbbj 10 | * [Baidu Cloud](https://pan.baidu.com/s/1x4t9hp1JIKKirI9s8Vsy1w) Fetchcode: aqaw && [Google Cloud](https://drive.google.com/file/d/1rStw0BybbGjARLfEEwcGuxuDv0oQoBFY/view?usp=sharing) 11 | 12 | ## Instance co-segmentation and co-saliency (published in TMM) 13 | * Here is our extended transaction paper https://ieeexplore.ieee.org/abstract/document/9337219/ 14 | * Instance co-saliency/segmentation maps: 15 | [Baidu Cloud](https://pan.baidu.com/s/1VxH2jX2d1oHxlKTYqtutQA) Fetchcodes: 05em 16 | * Instance evaluation codes (Matlab): 17 | [Baidu Cloud](https://pan.baidu.com/s/1QiC5GFcJ8YmeA7gb7y57sQ) Fetchcodes: v1up 18 | * Google Cloud: 19 | [link](https://drive.google.com/file/d/1Fn2zRC5agG_oTo1kz1_nU_aZGLgHGC23/view?usp=sharing) 20 | 21 | ## Citation 22 | If you use this code, please cite our paper: 23 | ``` 24 | @inproceedings{zhang2020adaptive, 25 | title={Adaptive graph convolutional network with attention graph clustering for co-saliency detection}, 26 | author={Zhang, Kaihua and Li, Tengpeng and Shen, Shiwen and Liu, Bo and Chen, Jin and Liu, Qingshan}, 27 | booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition}, 28 | pages={9050--9059}, 29 | year={2020} 30 | } 31 | @article{li2021image, 32 | title={Image Co-saliency Detection and Instance Co-segmentation using Attention Graph Clustering based Graph Convolutional Network}, 33 | author={Li, Tengpeng and Zhang, Kaihua and Shen, Shiwen and Liu, Bo and Liu, Qingshan and Li, Zhu}, 34 | journal={IEEE Transactions on Multimedia}, 35 | year={2021}, 36 | publisher={IEEE} 37 | } 38 | ``` 39 | 40 | -------------------------------------------------------------------------------- /model3/default.py: -------------------------------------------------------------------------------- 1 | 2 | # ------------------------------------------------------------------------------ 3 | # Copyright (c) Microsoft 4 | # Licensed under the MIT License. 5 | # Written by Ke Sun (sunk@mail.ustc.edu.cn) 6 | # ------------------------------------------------------------------------------ 7 | 8 | from __future__ import absolute_import 9 | from __future__ import division 10 | from __future__ import print_function 11 | 12 | import os 13 | 14 | from yacs.config import CfgNode as CN 15 | 16 | 17 | _C = CN() 18 | # 19 | # # common params for NETWORK 20 | # _C.MODEL = CN() 21 | # _C.MODEL.NAME = 'seg_hrnet' 22 | # _C.MODEL.PRETRAINED = '' 23 | # _C.MODEL.EXTRA = CN(new_allowed=True) 24 | # 25 | # # stage2 26 | # _C.MODEL.EXTRA.STAGE2 = CN() 27 | # _C.MODEL.EXTRA.STAGE2.NUM_MODULES = 1 28 | # _C.MODEL.EXTRA.STAGE2.NUM_BRANCHES = 2 29 | # _C.MODEL.EXTRA.STAGE2.BLOCK = 'BASIC' 30 | # _C.MODEL.EXTRA.STAGE2.NUM_BLOCKS = [4, 4] 31 | # _C.MODEL.EXTRA.STAGE2.NUM_CHANNELS = [48, 96] 32 | # _C.MODEL.EXTRA.STAGE2.FUSE_METHOD = 'SUM' 33 | # 34 | # # stage3 35 | # _C.MODEL.EXTRA.STAGE3 = CN() 36 | # _C.MODEL.EXTRA.STAGE3.NUM_MODULES = 4 37 | # _C.MODEL.EXTRA.STAGE3.NUM_BRANCHES = 3 38 | # _C.MODEL.EXTRA.STAGE3.BLOCK = 'BASIC' 39 | # _C.MODEL.EXTRA.STAGE3.NUM_BLOCKS = [4, 4, 4] 40 | # _C.MODEL.EXTRA.STAGE3.NUM_CHANNELS = [48, 96, 192] 41 | # _C.MODEL.EXTRA.STAGE3.FUSE_METHOD = 'SUM' 42 | # 43 | # # stage4 44 | # _C.MODEL.EXTRA.STAGE4 = CN() 45 | # _C.MODEL.EXTRA.STAGE4.NUM_MODULES = 3 46 | # _C.MODEL.EXTRA.STAGE4.NUM_BRANCHES = 4 47 | # _C.MODEL.EXTRA.STAGE4.BLOCK = 'BASIC' 48 | # _C.MODEL.EXTRA.STAGE4.NUM_BLOCKS = [4, 4, 4, 4] 49 | # _C.MODEL.EXTRA.STAGE4.NUM_CHANNELS = [48, 96, 192, 384] 50 | # _C.MODEL.EXTRA.STAGE4.FUSE_METHOD = 'SUM' 51 | 52 | _C.MODEL = CN() 53 | _C.MODEL.EXTRA = CN(new_allowed=True) 54 | _C.MODEL.PRETRAINED_LAYERS = ['*'] 55 | _C.MODEL.STEM_INPLANES = 64 56 | _C.MODEL.FINAL_CONV_KERNEL = 1 57 | _C.MODEL.WITH_HEAD = True 58 | 59 | _C.MODEL.EXTRA.STAGE2 = CN() 60 | _C.MODEL.EXTRA.STAGE2.NUM_MODULES = 1 61 | _C.MODEL.EXTRA.STAGE2.NUM_BRANCHES = 2 62 | _C.MODEL.EXTRA.STAGE2.NUM_BLOCKS = [4, 4] 63 | _C.MODEL.EXTRA.STAGE2.NUM_CHANNELS = [48, 96] 64 | _C.MODEL.EXTRA.STAGE2.BLOCK = 'BASIC' 65 | _C.MODEL.EXTRA.STAGE2.FUSE_METHOD = 'SUM' 66 | 67 | _C.MODEL.EXTRA.STAGE3 = CN() 68 | _C.MODEL.EXTRA.STAGE3.NUM_MODULES = 4 69 | _C.MODEL.EXTRA.STAGE3.NUM_BRANCHES = 3 70 | _C.MODEL.EXTRA.STAGE3.NUM_BLOCKS = [4, 4, 4] 71 | _C.MODEL.EXTRA.STAGE3.NUM_CHANNELS = [48, 96, 192] 72 | _C.MODEL.EXTRA.STAGE3.BLOCK = 'BASIC' 73 | _C.MODEL.EXTRA.STAGE3.FUSE_METHOD = 'SUM' 74 | 75 | _C.MODEL.EXTRA.STAGE4 = CN() 76 | _C.MODEL.EXTRA.STAGE4.NUM_MODULES = 3 77 | _C.MODEL.EXTRA.STAGE4.NUM_BRANCHES = 4 78 | _C.MODEL.EXTRA.STAGE4.NUM_BLOCKS = [4, 4, 4, 4] 79 | _C.MODEL.EXTRA.STAGE4.NUM_CHANNELS = [48, 96, 192, 384] 80 | _C.MODEL.EXTRA.STAGE4.BLOCK = 'BASIC' 81 | _C.MODEL.EXTRA.STAGE4.FUSE_METHOD = 'SUM' 82 | 83 | 84 | 85 | -------------------------------------------------------------------------------- /loss.py: -------------------------------------------------------------------------------- 1 | from torch import nn 2 | import torch.nn.functional as F 3 | import torch 4 | 5 | class Bce_Loss(nn.Module): 6 | def __init__(self): 7 | super(Bce_Loss, self).__init__() 8 | 9 | def forward(self, x, label): 10 | loss = F.binary_cross_entropy(x, label) 11 | return loss 12 | 13 | class Weighed_Bce_Loss(nn.Module): 14 | def __init__(self): 15 | super(Weighed_Bce_Loss, self).__init__() 16 | 17 | def forward(self, x, label): 18 | x = x.view(-1, 1, x.shape[1], x.shape[2]) 19 | label = label.view(-1, 1, label.shape[1], label.shape[2]) 20 | label_t = (label == 1).float() 21 | label_f = (label == 0).float() 22 | p = torch.sum(label_t) / (torch.sum(label_t) + torch.sum(label_f)) 23 | w = torch.zeros_like(label) 24 | w[label == 1] = p 25 | w[label == 0] = 1 - p 26 | loss = F.binary_cross_entropy(x, label, weight=w) 27 | return loss 28 | 29 | 30 | class OhemCrossEntropy(nn.Module): 31 | def __init__(self, thres, min_kept=100000): 32 | super(OhemCrossEntropy, self).__init__() 33 | self.thresh = thres 34 | self.min_kept = max(1, min_kept) 35 | 36 | def forward(self, x, label): 37 | pixel_losses = F.binary_cross_entropy(x, label, reduction='none').contiguous().view(-1) 38 | pred, ind = x.contiguous().view(-1).sort() 39 | pixel_losses = pixel_losses[pred < self.thresh] 40 | return pixel_losses.mean() 41 | 42 | class Cls_Loss(nn.Module): 43 | def __init__(self): 44 | super(Cls_Loss, self).__init__() 45 | 46 | def forward(self, x, label): 47 | loss = F.binary_cross_entropy(x, label) 48 | return loss 49 | 50 | class S_Loss(nn.Module): 51 | def __init__(self): 52 | super(S_Loss, self).__init__() 53 | 54 | def forward(self, x, label): 55 | loss = F.smooth_l1_loss(x, label) 56 | return loss 57 | 58 | class adj_Loss(nn.Module): 59 | def __init__(self): 60 | super(adj_Loss, self).__init__() 61 | 62 | def forward(self, x, label): 63 | x = x.view(-1, 1, x.shape[1], x.shape[2]) 64 | label = label.view(-1, 1, label.shape[1], label.shape[2]) 65 | loss = F.binary_cross_entropy(x, label) 66 | return loss 67 | ############################################# 68 | ### /home/litengpeng/CODE/co-segmentation/MaCoSNet-pytorch-master/model/ 69 | class Loss2(nn.Module): 70 | def __init__(self): 71 | super(Loss2, self).__init__() 72 | self.loss_wbce = Weighed_Bce_Loss() 73 | self.loss_s = S_Loss() 74 | self.w_bce = 1 75 | self.w_smooth = 1 76 | 77 | def forward(self, x, label): 78 | m_loss = self.loss_wbce(x, label) * self.w_bce 79 | s_loss = self.loss_s(x, label) * self.w_smooth 80 | loss = m_loss + s_loss 81 | 82 | return loss, m_loss, s_loss 83 | 84 | if __name__ == '__main__': 85 | x = torch.rand(4, 1, 224, 224) 86 | label = torch.zeros(4, 1, 224, 224) 87 | label[:, :, 30:80, 30:80] = 1 88 | loss = OhemCrossEntropy() 89 | l = loss(x, label) -------------------------------------------------------------------------------- /misc.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import os 3 | 4 | import pydensecrf.densecrf as dcrf 5 | import cv2 6 | 7 | class AvgMeter(object): 8 | def __init__(self): 9 | self.reset() 10 | 11 | def reset(self): 12 | self.val = 0 13 | self.avg = 0 14 | self.sum = 0 15 | self.count = 0 16 | 17 | def update(self, val, n=1): 18 | self.val = val 19 | self.sum += val * n 20 | self.count += n 21 | self.avg = self.sum / self.count 22 | 23 | 24 | def check_mkdir(dir_name): 25 | if not os.path.exists(dir_name): 26 | os.mkdir(dir_name) 27 | 28 | 29 | def cal_precision_recall_mae(prediction, gt): 30 | # input should be np array with data type uint8 31 | [w,h]=gt.shape 32 | prediction=cv2.resize(prediction, (h,w)) 33 | assert prediction.dtype == np.uint8 34 | assert gt.dtype == np.uint8 35 | assert prediction.shape == gt.shape 36 | 37 | eps = 1e-4 38 | 39 | prediction = prediction / 255. 40 | gt = gt / 255. 41 | 42 | mae = np.mean(np.abs(prediction - gt)) 43 | 44 | hard_gt = np.zeros(prediction.shape) 45 | hard_gt[gt > 0.5] = 1 46 | t = np.sum(hard_gt) 47 | 48 | precision, recall = [], [] 49 | # calculating precision and recall at 255 different binarizing thresholds 50 | for threshold in range(256): 51 | threshold = threshold / 255. 52 | 53 | hard_prediction = np.zeros(prediction.shape) 54 | hard_prediction[prediction > threshold] = 1 55 | 56 | tp = np.sum(hard_prediction * hard_gt) 57 | p = np.sum(hard_prediction) 58 | 59 | precision.append((tp + eps) / (p + eps)) 60 | recall.append((tp + eps) / (t + eps)) 61 | 62 | return precision, recall, mae 63 | 64 | 65 | def cal_fmeasure(precision, recall): 66 | assert len(precision) == 256 67 | assert len(recall) == 256 68 | beta_square = 0.3 69 | max_fmeasure = max([(1 + beta_square) * p * r / (beta_square * p + r) for p, r in zip(precision, recall)]) 70 | 71 | return max_fmeasure 72 | 73 | 74 | # codes of this function are borrowed from https://github.com/Andrew-Qibin/dss_crf 75 | def crf_refine(img, annos): 76 | def _sigmoid(x): 77 | return 1 / (1 + np.exp(-x)) 78 | 79 | assert img.dtype == np.uint8 80 | assert annos.dtype == np.uint8 81 | assert img.shape[:2] == annos.shape 82 | 83 | # img and annos should be np array with data type uint8 84 | 85 | EPSILON = 1e-8 86 | 87 | M = 2 # salient or not 88 | tau = 1.05 89 | # Setup the CRF model 90 | d = dcrf.DenseCRF2D(img.shape[1], img.shape[0], M) 91 | 92 | anno_norm = annos / 255. 93 | 94 | n_energy = -np.log((1.0 - anno_norm + EPSILON)) / (tau * _sigmoid(1 - anno_norm)) 95 | p_energy = -np.log(anno_norm + EPSILON) / (tau * _sigmoid(anno_norm)) 96 | 97 | U = np.zeros((M, img.shape[0] * img.shape[1]), dtype='float32') 98 | U[0, :] = n_energy.flatten() 99 | U[1, :] = p_energy.flatten() 100 | 101 | d.setUnaryEnergy(U) 102 | 103 | d.addPairwiseGaussian(sxy=3, compat=3) 104 | d.addPairwiseBilateral(sxy=60, srgb=5, rgbim=img, compat=5) 105 | 106 | # Do the inference 107 | infer = np.array(d.inference(1)).astype('float32') 108 | res = infer[1, :] 109 | 110 | res = res * 255 111 | res = res.reshape(img.shape[:2]) 112 | return res.astype('uint8') 113 | -------------------------------------------------------------------------------- /evalcodes-eccv18/EvalSaliency/EvalResultClassResult.m: -------------------------------------------------------------------------------- 1 | function [MAE, SMeasure, subAP, subAUC, f_f, tpr, fpr, pre, F_tpr, F_pre, WFmeasure] = EvalResultClassResult(ResultDir, GTPath, GTMaskExt) 2 | soddir2 = dir(GTPath); 3 | soddir2(1:2) = []; 4 | pre=[]; 5 | tpr=[]; 6 | fpr=[]; 7 | F_tpr=[]; 8 | F_pre=[]; 9 | MAE0=[]; 10 | subAP = zeros(1, length(soddir2), 'single'); 11 | subAUC = zeros(1, length(soddir2), 'single'); 12 | f_f = zeros(1, length(soddir2), 'single'); 13 | SMeasure = zeros(1, length(soddir2), 'single'); 14 | WFmeasure = zeros(1, length(soddir2), 'single'); 15 | for j = 1:length(soddir2) 16 | soddirnamewhole2=strcat(GTPath,soddir2(j).name,'/'); 17 | gtfile= dir(fullfile(soddirnamewhole2,GTMaskExt)); 18 | soddirnamewhole=strcat(ResultDir,soddir2(j).name,'/'); 19 | 20 | sub_pre=[]; 21 | sub_tpr=[]; 22 | sub_fpr=[]; 23 | f_tpr=[]; 24 | f_pre=[]; 25 | MAE1=[]; 26 | % 27 | % 28 | TempSMeasure = zeros(1, length(gtfile)); 29 | TempWFmeasure = zeros(length(gtfile),1, 'single'); 30 | for i=1:length(gtfile) 31 | gt=imread([soddirnamewhole2 gtfile(i).name(1:end-3) 'png']); 32 | gt = logical(gt(:,:,1)); 33 | if min(gt(:))==1 34 | continue 35 | end 36 | img=imread([soddirnamewhole gtfile(i).name(1:end-3) 'png']); 37 | TempSMeasure(i) = EvalStructureMeasure(img, gt); 38 | img=mat2gray(imresize(img,size(gt))); 39 | if size(img,3)>1 40 | img=img(:,:,1); 41 | end 42 | % [Precision,TPR, FPR] = QXL_ROC( img,gt , 100 ); 43 | [Precision,TPR, FPR] = EvalROC( img,gt , 100 ); 44 | pre = cat(1, pre, Precision); 45 | tpr = cat(1, tpr, TPR); 46 | fpr = cat(1, fpr, FPR); 47 | 48 | sub_pre= cat(1, sub_pre, Precision); 49 | sub_tpr = cat(1, sub_tpr, TPR); 50 | sub_fpr = cat(1, sub_fpr, FPR); 51 | TempWFmeasure(i) = WFb(img,gt); 52 | % % 53 | [~ ,TPR ,Precision] = Fmeasure( img, gt ); 54 | % % 55 | mae = mean2(abs(double(logical(gt)) - img)); 56 | 57 | MAE1 = [MAE1;mae]; 58 | f_tpr = cat(2, f_tpr, TPR); 59 | f_pre = cat(2, f_pre, Precision); 60 | 61 | F_tpr = cat(2, F_tpr, TPR); 62 | F_pre = cat(2, F_pre, Precision); 63 | 64 | end 65 | sub_T=mean(sub_tpr,1); 66 | sub_F=mean(sub_fpr,1); 67 | sub_P=mean(sub_pre,1); 68 | subAP(j) = -trapz(sub_T, sub_P); 69 | subAUC(j) = -trapz(sub_F, sub_T); 70 | f_f(j) = mean(f_tpr)*mean(f_pre)/(mean(f_tpr)+0.3*mean(f_pre))*1.3; 71 | SMeasure(j) = mean(TempSMeasure); 72 | WFmeasure(j) = mean(TempWFmeasure); 73 | MAE0=[MAE0;MAE1]; 74 | end 75 | MAE=mean(MAE0); 76 | end 77 | 78 | function [Q]= WFb(FG,GT) 79 | % WFb Compute the Weighted F-beta measure (as proposed in "How to Evaluate 80 | % Foreground Maps?" [Margolin et. al - CVPR'14]) 81 | % Usage: 82 | % Q = FbW(FG,GT) 83 | % Input: 84 | % FG - Binary/Non binary foreground map with values in the range [0 1]. Type: double. 85 | % GT - Binary ground truth. Type: logical. 86 | % Output: 87 | % Q - The Weighted F-beta score 88 | 89 | %Check input 90 | if (~isa( FG, 'double' )) 91 | error('FG should be of type: double'); 92 | end 93 | if ((max(FG(:))>1) || min(FG(:))<0) 94 | error('FG should be in the range of [0 1]'); 95 | end 96 | if (~islogical(GT)) 97 | error('GT should be of type: logical'); 98 | end 99 | 100 | dGT = double(GT); %Use double for computations. 101 | 102 | 103 | E = abs(FG-dGT); 104 | % [Ef, Et, Er] = deal(abs(FG-GT)); 105 | 106 | [Dst,IDXT] = bwdist(dGT); 107 | %Pixel dependency 108 | K = fspecial('gaussian',7,5); 109 | Et = E; 110 | Et(~GT)=Et(IDXT(~GT)); %To deal correctly with the edges of the foreground region 111 | EA = imfilter(Et,K); 112 | MIN_E_EA = E; 113 | MIN_E_EA(GT & EA= 1: 115 | d_loss = -1 * torch.log(2 - spatial_s_d) 116 | else: 117 | d_loss = -1 * torch.log(spatial_s_d) 118 | 119 | all_loss = 50 * d_loss + f_spa_loss 120 | # if iter%20==0: 121 | # print('iter: [%.4f], loss: [%.4f], dloss:[%.4f], floss: [%.4f]' %(iter, all_loss, d_loss, f_spa_loss)) 122 | 123 | spatial_s_optimizer.zero_grad() 124 | all_loss.backward() 125 | spatial_s_optimizer.step() 126 | 127 | result_map = spatial_s.data.view(5, 1, 7, 7) 128 | 129 | if i == 0: 130 | spa_mask = result_map 131 | else: 132 | spa_mask = torch.cat(([spa_mask, result_map]), dim=0) 133 | 134 | return spa_mask 135 | ##################### unsupervised masks 136 | ##################### unsupervised masks 137 | def row_normalize(mx): 138 | """Row-normalize sparse matrix""" 139 | rowsum = torch.sum(mx, dim=1) 140 | r_inv = 1 / (rowsum + 1e-10) 141 | r_mat_inv = torch.diag(r_inv) 142 | mx = torch.mm(r_mat_inv, mx) 143 | return mx 144 | def unsqz_fea(dim4_data): 145 | split_data = torch.split(dim4_data, 5, dim=0) 146 | for i in range(len(split_data)): 147 | if i == 0: 148 | dim5_data = split_data[i].unsqueeze(dim=0) 149 | else: 150 | dim5_data = torch.cat((dim5_data, split_data[i].unsqueeze(dim=0)), dim=0) 151 | return dim5_data 152 | 153 | def sqz_fea(dim5_data): 154 | if dim5_data.size(1) == 1: 155 | return dim5_data.squeeze() 156 | else: 157 | b = dim5_data.size(0) 158 | for i in range(b): 159 | if i == 0: 160 | new_dim4_data = dim5_data[i, :, :, :, :] 161 | else: 162 | new_dim4_data = torch.cat((new_dim4_data, dim5_data[i, :, :, :, :]), dim=0) 163 | return new_dim4_data 164 | ################################ 165 | class GraphConvolution(Module): 166 | """ 167 | Simple GCN layer, similar to https://arxiv.org/abs/1609.02907 168 | """ 169 | 170 | def __init__(self, in_features, out_features, bias=True): 171 | super(GraphConvolution, self).__init__() 172 | self.in_features = in_features 173 | self.out_features = out_features 174 | self.weight = Parameter(torch.FloatTensor(in_features, out_features)) 175 | if bias: 176 | self.bias = Parameter(torch.FloatTensor(out_features)) 177 | else: 178 | self.register_parameter('bias', None) 179 | self.reset_parameters() 180 | 181 | def reset_parameters(self): 182 | stdv = 1. / math.sqrt(self.weight.size(1)) 183 | self.weight.data.uniform_(-stdv, stdv) 184 | if self.bias is not None: 185 | self.bias.data.uniform_(-stdv, stdv) 186 | 187 | def forward(self, input, adj): 188 | support = torch.mm(input, self.weight) 189 | output = torch.mm(adj, support) 190 | if self.bias is not None: 191 | return output + self.bias 192 | else: 193 | return output 194 | ################# 195 | decoder_archs = { 196 | 'd16': [336, 'd128', 128, 128, 'd64', 64, 64, 'c1'] 197 | } 198 | 199 | def make_decoder_layers(cfg, in_channels, batch_norm=True): 200 | layers = [] 201 | for v in cfg: 202 | if type(v) is str: 203 | if v[0] == 'd': 204 | v = int(v[1:]) 205 | convtrans2d = nn.ConvTranspose2d(in_channels, v, kernel_size=4, stride=2, padding=1) 206 | if batch_norm: 207 | layers += [convtrans2d, nn.BatchNorm2d(v), nn.ReLU()] 208 | else: 209 | layers += [convtrans2d, nn.ReLU()] 210 | in_channels = v 211 | elif v[0] == 'c': 212 | v = int(v[1:]) 213 | layers += [nn.Conv2d(in_channels, v, kernel_size=3, padding=1)] 214 | else: 215 | conv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1) 216 | if batch_norm: 217 | layers += [conv2d, nn.BatchNorm2d(v), nn.ReLU()] 218 | else: 219 | layers += [conv2d, nn.ReLU()] 220 | in_channels = v 221 | return nn.Sequential(*layers) 222 | class DOCSDecoderNet(nn.Module): 223 | def __init__(self, features): 224 | super(DOCSDecoderNet, self).__init__() 225 | self.features = features 226 | 227 | def forward(self, x): 228 | return self.features(x) -------------------------------------------------------------------------------- /model3/cls_hrnet.py: -------------------------------------------------------------------------------- 1 | # ------------------------------------------------------------------------------ 2 | # Copyright (c) Microsoft 3 | # Licensed under the MIT License. 4 | # Written by Bin Xiao (Bin.Xiao@microsoft.com) 5 | # Modified by Ke Sun (sunk@mail.ustc.edu.cn) 6 | # ------------------------------------------------------------------------------ 7 | 8 | from __future__ import absolute_import 9 | from __future__ import division 10 | from __future__ import print_function 11 | 12 | import os 13 | import logging 14 | import functools 15 | 16 | import numpy as np 17 | 18 | import torch 19 | import torch.nn as nn 20 | import torch._utils 21 | import torch.nn.functional as F 22 | 23 | BN_MOMENTUM = 0.1 24 | logger = logging.getLogger(__name__) 25 | 26 | 27 | def conv3x3(in_planes, out_planes, stride=1): 28 | """3x3 convolution with padding""" 29 | return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, 30 | padding=1, bias=False) 31 | 32 | 33 | class BasicBlock(nn.Module): 34 | expansion = 1 35 | 36 | def __init__(self, inplanes, planes, stride=1, downsample=None): 37 | super(BasicBlock, self).__init__() 38 | self.conv1 = conv3x3(inplanes, planes, stride) 39 | self.bn1 = nn.BatchNorm2d(planes, momentum=BN_MOMENTUM) 40 | self.relu = nn.ReLU(inplace=True) 41 | self.conv2 = conv3x3(planes, planes) 42 | self.bn2 = nn.BatchNorm2d(planes, momentum=BN_MOMENTUM) 43 | self.downsample = downsample 44 | self.stride = stride 45 | 46 | def forward(self, x): 47 | residual = x 48 | 49 | out = self.conv1(x) 50 | out = self.bn1(out) 51 | out = self.relu(out) 52 | 53 | out = self.conv2(out) 54 | out = self.bn2(out) 55 | 56 | if self.downsample is not None: 57 | residual = self.downsample(x) 58 | 59 | out += residual 60 | out = self.relu(out) 61 | 62 | return out 63 | 64 | 65 | class Bottleneck(nn.Module): 66 | expansion = 4 67 | 68 | def __init__(self, inplanes, planes, stride=1, downsample=None): 69 | super(Bottleneck, self).__init__() 70 | self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False) 71 | self.bn1 = nn.BatchNorm2d(planes, momentum=BN_MOMENTUM) 72 | self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, 73 | padding=1, bias=False) 74 | self.bn2 = nn.BatchNorm2d(planes, momentum=BN_MOMENTUM) 75 | self.conv3 = nn.Conv2d(planes, planes * self.expansion, kernel_size=1, 76 | bias=False) 77 | self.bn3 = nn.BatchNorm2d(planes * self.expansion, 78 | momentum=BN_MOMENTUM) 79 | self.relu = nn.ReLU(inplace=True) 80 | self.downsample = downsample 81 | self.stride = stride 82 | 83 | def forward(self, x): 84 | residual = x 85 | 86 | out = self.conv1(x) 87 | out = self.bn1(out) 88 | out = self.relu(out) 89 | 90 | out = self.conv2(out) 91 | out = self.bn2(out) 92 | out = self.relu(out) 93 | 94 | out = self.conv3(out) 95 | out = self.bn3(out) 96 | 97 | if self.downsample is not None: 98 | residual = self.downsample(x) 99 | 100 | out += residual 101 | out = self.relu(out) 102 | 103 | return out 104 | 105 | 106 | class HighResolutionModule(nn.Module): 107 | def __init__(self, num_branches, blocks, num_blocks, num_inchannels, 108 | num_channels, fuse_method, multi_scale_output=True): 109 | super(HighResolutionModule, self).__init__() 110 | self._check_branches( 111 | num_branches, blocks, num_blocks, num_inchannels, num_channels) 112 | 113 | self.num_inchannels = num_inchannels 114 | self.fuse_method = fuse_method 115 | self.num_branches = num_branches 116 | 117 | self.multi_scale_output = multi_scale_output 118 | 119 | self.branches = self._make_branches( 120 | num_branches, blocks, num_blocks, num_channels) 121 | self.fuse_layers = self._make_fuse_layers() 122 | self.relu = nn.ReLU(False) 123 | 124 | def _check_branches(self, num_branches, blocks, num_blocks, 125 | num_inchannels, num_channels): 126 | if num_branches != len(num_blocks): 127 | error_msg = 'NUM_BRANCHES({}) <> NUM_BLOCKS({})'.format( 128 | num_branches, len(num_blocks)) 129 | logger.error(error_msg) 130 | raise ValueError(error_msg) 131 | 132 | if num_branches != len(num_channels): 133 | error_msg = 'NUM_BRANCHES({}) <> NUM_CHANNELS({})'.format( 134 | num_branches, len(num_channels)) 135 | logger.error(error_msg) 136 | raise ValueError(error_msg) 137 | 138 | if num_branches != len(num_inchannels): 139 | error_msg = 'NUM_BRANCHES({}) <> NUM_INCHANNELS({})'.format( 140 | num_branches, len(num_inchannels)) 141 | logger.error(error_msg) 142 | raise ValueError(error_msg) 143 | 144 | def _make_one_branch(self, branch_index, block, num_blocks, num_channels, 145 | stride=1): 146 | downsample = None 147 | if stride != 1 or \ 148 | self.num_inchannels[branch_index] != num_channels[branch_index] * block.expansion: 149 | downsample = nn.Sequential( 150 | nn.Conv2d(self.num_inchannels[branch_index], 151 | num_channels[branch_index] * block.expansion, 152 | kernel_size=1, stride=stride, bias=False), 153 | nn.BatchNorm2d(num_channels[branch_index] * block.expansion, 154 | momentum=BN_MOMENTUM), 155 | ) 156 | 157 | layers = [] 158 | layers.append(block(self.num_inchannels[branch_index], 159 | num_channels[branch_index], stride, downsample)) 160 | self.num_inchannels[branch_index] = \ 161 | num_channels[branch_index] * block.expansion 162 | for i in range(1, num_blocks[branch_index]): 163 | layers.append(block(self.num_inchannels[branch_index], 164 | num_channels[branch_index])) 165 | 166 | return nn.Sequential(*layers) 167 | 168 | def _make_branches(self, num_branches, block, num_blocks, num_channels): 169 | branches = [] 170 | 171 | for i in range(num_branches): 172 | branches.append( 173 | self._make_one_branch(i, block, num_blocks, num_channels)) 174 | 175 | return nn.ModuleList(branches) 176 | 177 | def _make_fuse_layers(self): 178 | if self.num_branches == 1: 179 | return None 180 | 181 | num_branches = self.num_branches 182 | num_inchannels = self.num_inchannels 183 | fuse_layers = [] 184 | for i in range(num_branches if self.multi_scale_output else 1): 185 | fuse_layer = [] 186 | for j in range(num_branches): 187 | if j > i: 188 | fuse_layer.append(nn.Sequential( 189 | nn.Conv2d(num_inchannels[j], 190 | num_inchannels[i], 191 | 1, 192 | 1, 193 | 0, 194 | bias=False), 195 | nn.BatchNorm2d(num_inchannels[i], 196 | momentum=BN_MOMENTUM), 197 | nn.Upsample(scale_factor=2**(j-i), mode='nearest'))) 198 | elif j == i: 199 | fuse_layer.append(None) 200 | else: 201 | conv3x3s = [] 202 | for k in range(i-j): 203 | if k == i - j - 1: 204 | num_outchannels_conv3x3 = num_inchannels[i] 205 | conv3x3s.append(nn.Sequential( 206 | nn.Conv2d(num_inchannels[j], 207 | num_outchannels_conv3x3, 208 | 3, 2, 1, bias=False), 209 | nn.BatchNorm2d(num_outchannels_conv3x3, 210 | momentum=BN_MOMENTUM))) 211 | else: 212 | num_outchannels_conv3x3 = num_inchannels[j] 213 | conv3x3s.append(nn.Sequential( 214 | nn.Conv2d(num_inchannels[j], 215 | num_outchannels_conv3x3, 216 | 3, 2, 1, bias=False), 217 | nn.BatchNorm2d(num_outchannels_conv3x3, 218 | momentum=BN_MOMENTUM), 219 | nn.ReLU(False))) 220 | fuse_layer.append(nn.Sequential(*conv3x3s)) 221 | fuse_layers.append(nn.ModuleList(fuse_layer)) 222 | 223 | return nn.ModuleList(fuse_layers) 224 | 225 | def get_num_inchannels(self): 226 | return self.num_inchannels 227 | 228 | def forward(self, x): 229 | if self.num_branches == 1: 230 | return [self.branches[0](x[0])] 231 | 232 | for i in range(self.num_branches): 233 | x[i] = self.branches[i](x[i]) 234 | 235 | x_fuse = [] 236 | for i in range(len(self.fuse_layers)): 237 | y = x[0] if i == 0 else self.fuse_layers[i][0](x[0]) 238 | for j in range(1, self.num_branches): 239 | if i == j: 240 | y = y + x[j] 241 | else: 242 | y = y + self.fuse_layers[i][j](x[j]) 243 | x_fuse.append(self.relu(y)) 244 | 245 | return x_fuse 246 | 247 | 248 | blocks_dict = { 249 | 'BASIC': BasicBlock, 250 | 'BOTTLENECK': Bottleneck 251 | } 252 | 253 | 254 | class HighResolutionNet(nn.Module): 255 | 256 | def __init__(self, cfg, **kwargs): 257 | super(HighResolutionNet, self).__init__() 258 | 259 | self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=2, padding=1, 260 | bias=False) 261 | self.bn1 = nn.BatchNorm2d(64, momentum=BN_MOMENTUM) 262 | self.conv2 = nn.Conv2d(64, 64, kernel_size=3, stride=2, padding=1, 263 | bias=False) 264 | self.bn2 = nn.BatchNorm2d(64, momentum=BN_MOMENTUM) 265 | self.relu = nn.ReLU(inplace=True) 266 | self.layer1 = self._make_layer(Bottleneck, 64, 64, 4) 267 | 268 | self.stage2_cfg = cfg['MODEL']['EXTRA']['STAGE2'] 269 | num_channels = self.stage2_cfg['NUM_CHANNELS'] 270 | block = blocks_dict[self.stage2_cfg['BLOCK']] 271 | num_channels = [ 272 | num_channels[i] * block.expansion for i in range(len(num_channels))] 273 | self.transition1 = self._make_transition_layer( 274 | [256], num_channels) 275 | self.stage2, pre_stage_channels = self._make_stage( 276 | self.stage2_cfg, num_channels) 277 | 278 | self.stage3_cfg = cfg['MODEL']['EXTRA']['STAGE3'] 279 | num_channels = self.stage3_cfg['NUM_CHANNELS'] 280 | block = blocks_dict[self.stage3_cfg['BLOCK']] 281 | num_channels = [ 282 | num_channels[i] * block.expansion for i in range(len(num_channels))] 283 | self.transition2 = self._make_transition_layer( 284 | pre_stage_channels, num_channels) 285 | self.stage3, pre_stage_channels = self._make_stage( 286 | self.stage3_cfg, num_channels) 287 | 288 | self.stage4_cfg = cfg['MODEL']['EXTRA']['STAGE4'] 289 | num_channels = self.stage4_cfg['NUM_CHANNELS'] 290 | block = blocks_dict[self.stage4_cfg['BLOCK']] 291 | num_channels = [ 292 | num_channels[i] * block.expansion for i in range(len(num_channels))] 293 | self.transition3 = self._make_transition_layer( 294 | pre_stage_channels, num_channels) 295 | self.stage4, pre_stage_channels = self._make_stage( 296 | self.stage4_cfg, num_channels, multi_scale_output=True) 297 | 298 | # Classification Head 299 | # self.incre_modules, self.downsamp_modules, \ 300 | # self.final_layer = self._make_head(pre_stage_channels) 301 | # 302 | # self.classifier = nn.Linear(2048, 1000) 303 | 304 | # def _make_head(self, pre_stage_channels): 305 | # head_block = Bottleneck 306 | # head_channels = [32, 64, 128, 256] 307 | # 308 | # # Increasing the #channels on each resolution 309 | # # from C, 2C, 4C, 8C to 128, 256, 512, 1024 310 | # incre_modules = [] 311 | # for i, channels in enumerate(pre_stage_channels): 312 | # incre_module = self._make_layer(head_block, 313 | # channels, 314 | # head_channels[i], 315 | # 1, 316 | # stride=1) 317 | # incre_modules.append(incre_module) 318 | # incre_modules = nn.ModuleList(incre_modules) 319 | # 320 | # # downsampling modules 321 | # downsamp_modules = [] 322 | # for i in range(len(pre_stage_channels)-1): 323 | # in_channels = head_channels[i] * head_block.expansion 324 | # out_channels = head_channels[i+1] * head_block.expansion 325 | # 326 | # downsamp_module = nn.Sequential( 327 | # nn.Conv2d(in_channels=in_channels, 328 | # out_channels=out_channels, 329 | # kernel_size=3, 330 | # stride=2, 331 | # padding=1), 332 | # nn.BatchNorm2d(out_channels, momentum=BN_MOMENTUM), 333 | # nn.ReLU(inplace=True) 334 | # ) 335 | # 336 | # downsamp_modules.append(downsamp_module) 337 | # downsamp_modules = nn.ModuleList(downsamp_modules) 338 | # 339 | # final_layer = nn.Sequential( 340 | # nn.Conv2d( 341 | # in_channels=head_channels[3] * head_block.expansion, 342 | # out_channels=2048, 343 | # kernel_size=1, 344 | # stride=1, 345 | # padding=0 346 | # ), 347 | # nn.BatchNorm2d(2048, momentum=BN_MOMENTUM), 348 | # nn.ReLU(inplace=True) 349 | # ) 350 | # 351 | # return incre_modules, downsamp_modules, final_layer 352 | 353 | def _make_transition_layer( 354 | self, num_channels_pre_layer, num_channels_cur_layer): 355 | num_branches_cur = len(num_channels_cur_layer) 356 | num_branches_pre = len(num_channels_pre_layer) 357 | 358 | transition_layers = [] 359 | for i in range(num_branches_cur): 360 | if i < num_branches_pre: 361 | if num_channels_cur_layer[i] != num_channels_pre_layer[i]: 362 | transition_layers.append(nn.Sequential( 363 | nn.Conv2d(num_channels_pre_layer[i], 364 | num_channels_cur_layer[i], 365 | 3, 366 | 1, 367 | 1, 368 | bias=False), 369 | nn.BatchNorm2d( 370 | num_channels_cur_layer[i], momentum=BN_MOMENTUM), 371 | nn.ReLU(inplace=True))) 372 | else: 373 | transition_layers.append(None) 374 | else: 375 | conv3x3s = [] 376 | for j in range(i+1-num_branches_pre): 377 | inchannels = num_channels_pre_layer[-1] 378 | outchannels = num_channels_cur_layer[i] \ 379 | if j == i-num_branches_pre else inchannels 380 | conv3x3s.append(nn.Sequential( 381 | nn.Conv2d( 382 | inchannels, outchannels, 3, 2, 1, bias=False), 383 | nn.BatchNorm2d(outchannels, momentum=BN_MOMENTUM), 384 | nn.ReLU(inplace=True))) 385 | transition_layers.append(nn.Sequential(*conv3x3s)) 386 | 387 | return nn.ModuleList(transition_layers) 388 | 389 | def _make_layer(self, block, inplanes, planes, blocks, stride=1): 390 | downsample = None 391 | if stride != 1 or inplanes != planes * block.expansion: 392 | downsample = nn.Sequential( 393 | nn.Conv2d(inplanes, planes * block.expansion, 394 | kernel_size=1, stride=stride, bias=False), 395 | nn.BatchNorm2d(planes * block.expansion, momentum=BN_MOMENTUM), 396 | ) 397 | 398 | layers = [] 399 | layers.append(block(inplanes, planes, stride, downsample)) 400 | inplanes = planes * block.expansion 401 | for i in range(1, blocks): 402 | layers.append(block(inplanes, planes)) 403 | 404 | return nn.Sequential(*layers) 405 | 406 | def _make_stage(self, layer_config, num_inchannels, 407 | multi_scale_output=True): 408 | num_modules = layer_config['NUM_MODULES'] 409 | num_branches = layer_config['NUM_BRANCHES'] 410 | num_blocks = layer_config['NUM_BLOCKS'] 411 | num_channels = layer_config['NUM_CHANNELS'] 412 | block = blocks_dict[layer_config['BLOCK']] 413 | fuse_method = layer_config['FUSE_METHOD'] 414 | 415 | modules = [] 416 | for i in range(num_modules): 417 | # multi_scale_output is only used last module 418 | if not multi_scale_output and i == num_modules - 1: 419 | reset_multi_scale_output = False 420 | else: 421 | reset_multi_scale_output = True 422 | 423 | modules.append( 424 | HighResolutionModule(num_branches, 425 | block, 426 | num_blocks, 427 | num_inchannels, 428 | num_channels, 429 | fuse_method, 430 | reset_multi_scale_output) 431 | ) 432 | num_inchannels = modules[-1].get_num_inchannels() 433 | 434 | return nn.Sequential(*modules), num_inchannels 435 | 436 | def forward(self, x): 437 | x = self.conv1(x) 438 | x = self.bn1(x) 439 | x = self.relu(x) 440 | x = self.conv2(x) 441 | x = self.bn2(x) 442 | x = self.relu(x) 443 | x = self.layer1(x) 444 | 445 | x_list = [] 446 | for i in range(self.stage2_cfg['NUM_BRANCHES']): 447 | if self.transition1[i] is not None: 448 | x_list.append(self.transition1[i](x)) 449 | else: 450 | x_list.append(x) 451 | y_list = self.stage2(x_list) 452 | 453 | x_list = [] 454 | for i in range(self.stage3_cfg['NUM_BRANCHES']): 455 | if self.transition2[i] is not None: 456 | x_list.append(self.transition2[i](y_list[-1])) 457 | else: 458 | x_list.append(y_list[i]) 459 | y_list = self.stage3(x_list) 460 | 461 | x_list = [] 462 | for i in range(self.stage4_cfg['NUM_BRANCHES']): 463 | if self.transition3[i] is not None: 464 | x_list.append(self.transition3[i](y_list[-1])) 465 | else: 466 | x_list.append(y_list[i]) 467 | y_list = self.stage4(x_list) 468 | 469 | # Classification Head 470 | # y = self.incre_modules[0](y_list[0]) 471 | # for i in range(len(self.downsamp_modules)): 472 | # y = self.incre_modules[i+1](y_list[i+1]) + \ 473 | # self.downsamp_modules[i](y) 474 | # 475 | # y = self.final_layer(y) 476 | # 477 | # y = F.avg_pool2d(y, kernel_size=y.size() 478 | # [2:]).view(y.size(0), -1) 479 | # 480 | # y = self.classifier(y) 481 | 482 | return y_list 483 | 484 | def init_weights(self, pretrained='',): 485 | logger.info('=> init weights from normal distribution') 486 | for m in self.modules(): 487 | if isinstance(m, nn.Conv2d): 488 | nn.init.kaiming_normal_( 489 | m.weight, mode='fan_out', nonlinearity='relu') 490 | elif isinstance(m, nn.BatchNorm2d): 491 | nn.init.constant_(m.weight, 1) 492 | nn.init.constant_(m.bias, 0) 493 | if os.path.isfile(pretrained): 494 | pretrained_dict = torch.load(pretrained) 495 | logger.info('=> loading pretrained model {}'.format(pretrained)) 496 | model_dict = self.state_dict() 497 | pretrained_dict = {k: v for k, v in pretrained_dict.items() 498 | if k in model_dict.keys()} 499 | for k, _ in pretrained_dict.items(): 500 | logger.info( 501 | '=> loading {} pretrained model {}'.format(k, pretrained)) 502 | model_dict.update(pretrained_dict) 503 | self.load_state_dict(model_dict) 504 | 505 | 506 | def get_cls_net(config, **kwargs): 507 | model = HighResolutionNet(config, **kwargs) 508 | model.init_weights() 509 | return model 510 | -------------------------------------------------------------------------------- /log_file/graph4_decoder_hrnet_agcm_stage3.txt: -------------------------------------------------------------------------------- 1 | lr: 2.500000e-05, epoch: [0/120000], all_loss: [0.6176], m_loss:[0.5038], s_loss:[0.1137] 2 | lr: 2.500000e-05, epoch: [0/120000], all_loss: [0.6874], m_loss:[0.5582], s_loss:[0.1291] 3 | iCoseg Cosal2015 f: [0.2043], m: [0.3577], f: [0.2308], m: [0.5049] 4 | lr: 2.500000e-05, epoch: [0/120000], all_loss: [0.0609], m_loss:[0.0425], s_loss:[0.0184] 5 | iCoseg Cosal2015 f: [0.8414], m: [0.0808], f: [0.8375], m: [0.0924] 6 | lr: 2.500000e-05, epoch: [0/120000], all_loss: [0.0267], m_loss:[0.0199], s_loss:[0.0067] 7 | iCoseg Cosal2015 f: [0.8405], m: [0.0818], f: [0.8358], m: [0.0942] 8 | lr: 2.500000e-05, epoch: [100/120000], all_loss: [0.1215], m_loss:[0.0774], s_loss:[0.0441] 9 | lr: 2.500000e-05, epoch: [200/120000], all_loss: [0.0329], m_loss:[0.0179], s_loss:[0.0150] 10 | lr: 2.500000e-05, epoch: [300/120000], all_loss: [0.0513], m_loss:[0.0313], s_loss:[0.0200] 11 | lr: 2.500000e-05, epoch: [400/120000], all_loss: [0.0552], m_loss:[0.0404], s_loss:[0.0148] 12 | lr: 2.500000e-05, epoch: [500/120000], all_loss: [0.0375], m_loss:[0.0285], s_loss:[0.0091] 13 | lr: 2.500000e-05, epoch: [600/120000], all_loss: [0.0579], m_loss:[0.0421], s_loss:[0.0157] 14 | lr: 2.500000e-05, epoch: [700/120000], all_loss: [0.0481], m_loss:[0.0305], s_loss:[0.0176] 15 | lr: 2.500000e-05, epoch: [800/120000], all_loss: [0.0520], m_loss:[0.0369], s_loss:[0.0151] 16 | lr: 2.500000e-05, epoch: [900/120000], all_loss: [0.0608], m_loss:[0.0406], s_loss:[0.0202] 17 | lr: 2.500000e-05, epoch: [0/150000], all_loss: [0.0500], m_loss:[0.0321], s_loss:[0.0179] 18 | iCoseg Cosal2015 f: [0.8418], m: [0.0800], f: [0.8386], m: [0.0908] 19 | lr: 2.500000e-05, epoch: [100/150000], all_loss: [0.0259], m_loss:[0.0171], s_loss:[0.0088] 20 | lr: 2.500000e-05, epoch: [200/150000], all_loss: [0.0520], m_loss:[0.0317], s_loss:[0.0203] 21 | lr: 2.500000e-05, epoch: [300/150000], all_loss: [0.1470], m_loss:[0.0901], s_loss:[0.0569] 22 | lr: 2.500000e-05, epoch: [400/150000], all_loss: [0.0382], m_loss:[0.0252], s_loss:[0.0130] 23 | lr: 2.500000e-05, epoch: [500/150000], all_loss: [0.0788], m_loss:[0.0537], s_loss:[0.0250] 24 | lr: 2.500000e-05, epoch: [600/150000], all_loss: [0.0669], m_loss:[0.0451], s_loss:[0.0218] 25 | lr: 2.500000e-05, epoch: [700/150000], all_loss: [0.1391], m_loss:[0.0863], s_loss:[0.0529] 26 | lr: 2.500000e-05, epoch: [800/150000], all_loss: [0.0371], m_loss:[0.0267], s_loss:[0.0103] 27 | lr: 2.500000e-05, epoch: [900/150000], all_loss: [0.1475], m_loss:[0.1009], s_loss:[0.0466] 28 | lr: 2.500000e-05, epoch: [1000/150000], all_loss: [0.0615], m_loss:[0.0425], s_loss:[0.0190] 29 | lr: 2.500000e-05, epoch: [1100/150000], all_loss: [0.0388], m_loss:[0.0241], s_loss:[0.0148] 30 | lr: 2.500000e-05, epoch: [1200/150000], all_loss: [0.0712], m_loss:[0.0464], s_loss:[0.0248] 31 | lr: 2.500000e-05, epoch: [1300/150000], all_loss: [0.0561], m_loss:[0.0325], s_loss:[0.0237] 32 | lr: 2.500000e-05, epoch: [1400/150000], all_loss: [0.0600], m_loss:[0.0388], s_loss:[0.0212] 33 | lr: 2.500000e-05, epoch: [1500/150000], all_loss: [0.3331], m_loss:[0.2330], s_loss:[0.1001] 34 | lr: 2.500000e-05, epoch: [1600/150000], all_loss: [0.0411], m_loss:[0.0268], s_loss:[0.0143] 35 | lr: 2.500000e-05, epoch: [1700/150000], all_loss: [0.0233], m_loss:[0.0139], s_loss:[0.0093] 36 | lr: 2.500000e-05, epoch: [1800/150000], all_loss: [0.0811], m_loss:[0.0577], s_loss:[0.0234] 37 | lr: 2.500000e-05, epoch: [1900/150000], all_loss: [0.0288], m_loss:[0.0189], s_loss:[0.0099] 38 | lr: 2.500000e-05, epoch: [2000/150000], all_loss: [0.0695], m_loss:[0.0475], s_loss:[0.0220] 39 | lr: 2.500000e-05, epoch: [2100/150000], all_loss: [0.0611], m_loss:[0.0436], s_loss:[0.0176] 40 | lr: 2.500000e-05, epoch: [2200/150000], all_loss: [0.0366], m_loss:[0.0255], s_loss:[0.0111] 41 | lr: 2.500000e-05, epoch: [2300/150000], all_loss: [0.0346], m_loss:[0.0243], s_loss:[0.0103] 42 | lr: 2.500000e-05, epoch: [2400/150000], all_loss: [0.0712], m_loss:[0.0520], s_loss:[0.0192] 43 | lr: 2.500000e-05, epoch: [2500/150000], all_loss: [0.1758], m_loss:[0.1387], s_loss:[0.0371] 44 | iCoseg Cosal2015 f: [0.8430], m: [0.0797], f: [0.8466], m: [0.0875] 45 | lr: 2.500000e-05, epoch: [2600/150000], all_loss: [0.0710], m_loss:[0.0480], s_loss:[0.0230] 46 | lr: 2.500000e-05, epoch: [2700/150000], all_loss: [0.0856], m_loss:[0.0547], s_loss:[0.0309] 47 | lr: 2.500000e-05, epoch: [2800/150000], all_loss: [0.0763], m_loss:[0.0416], s_loss:[0.0347] 48 | lr: 2.500000e-05, epoch: [2900/150000], all_loss: [0.0233], m_loss:[0.0164], s_loss:[0.0069] 49 | lr: 2.500000e-05, epoch: [3000/150000], all_loss: [0.1584], m_loss:[0.1085], s_loss:[0.0499] 50 | lr: 2.500000e-05, epoch: [3100/150000], all_loss: [0.0729], m_loss:[0.0493], s_loss:[0.0236] 51 | lr: 2.500000e-05, epoch: [3200/150000], all_loss: [0.0355], m_loss:[0.0193], s_loss:[0.0161] 52 | lr: 2.500000e-05, epoch: [3300/150000], all_loss: [0.1527], m_loss:[0.1114], s_loss:[0.0413] 53 | lr: 2.500000e-05, epoch: [3400/150000], all_loss: [0.0303], m_loss:[0.0201], s_loss:[0.0102] 54 | lr: 2.500000e-05, epoch: [3500/150000], all_loss: [0.0416], m_loss:[0.0249], s_loss:[0.0167] 55 | lr: 2.500000e-05, epoch: [3600/150000], all_loss: [0.0537], m_loss:[0.0328], s_loss:[0.0209] 56 | lr: 2.500000e-05, epoch: [3700/150000], all_loss: [0.0712], m_loss:[0.0470], s_loss:[0.0242] 57 | lr: 2.500000e-05, epoch: [3800/150000], all_loss: [0.0374], m_loss:[0.0207], s_loss:[0.0167] 58 | lr: 2.500000e-05, epoch: [3900/150000], all_loss: [0.0688], m_loss:[0.0466], s_loss:[0.0222] 59 | lr: 2.500000e-05, epoch: [4000/150000], all_loss: [0.0571], m_loss:[0.0357], s_loss:[0.0214] 60 | lr: 2.500000e-05, epoch: [4100/150000], all_loss: [0.0731], m_loss:[0.0423], s_loss:[0.0307] 61 | lr: 2.500000e-05, epoch: [4200/150000], all_loss: [0.0444], m_loss:[0.0281], s_loss:[0.0163] 62 | lr: 2.500000e-05, epoch: [4300/150000], all_loss: [0.0388], m_loss:[0.0228], s_loss:[0.0159] 63 | lr: 2.500000e-05, epoch: [4400/150000], all_loss: [0.0749], m_loss:[0.0469], s_loss:[0.0280] 64 | lr: 2.500000e-05, epoch: [4500/150000], all_loss: [0.0526], m_loss:[0.0383], s_loss:[0.0144] 65 | lr: 2.500000e-05, epoch: [4600/150000], all_loss: [0.1015], m_loss:[0.0662], s_loss:[0.0353] 66 | lr: 2.500000e-05, epoch: [4700/150000], all_loss: [0.0996], m_loss:[0.0612], s_loss:[0.0384] 67 | lr: 2.500000e-05, epoch: [4800/150000], all_loss: [0.0836], m_loss:[0.0645], s_loss:[0.0191] 68 | lr: 2.500000e-05, epoch: [4900/150000], all_loss: [0.0270], m_loss:[0.0139], s_loss:[0.0131] 69 | lr: 2.500000e-05, epoch: [5000/150000], all_loss: [0.0435], m_loss:[0.0246], s_loss:[0.0190] 70 | iCoseg Cosal2015 f: [0.8397], m: [0.0793], f: [0.8476], m: [0.0874] 71 | lr: 2.500000e-05, epoch: [5100/150000], all_loss: [0.0507], m_loss:[0.0333], s_loss:[0.0174] 72 | lr: 2.500000e-05, epoch: [5200/150000], all_loss: [0.0522], m_loss:[0.0292], s_loss:[0.0230] 73 | lr: 2.500000e-05, epoch: [5300/150000], all_loss: [0.0700], m_loss:[0.0445], s_loss:[0.0256] 74 | lr: 2.500000e-05, epoch: [5400/150000], all_loss: [0.0317], m_loss:[0.0202], s_loss:[0.0115] 75 | lr: 2.500000e-05, epoch: [5500/150000], all_loss: [0.1082], m_loss:[0.0678], s_loss:[0.0404] 76 | lr: 2.500000e-05, epoch: [5600/150000], all_loss: [0.0488], m_loss:[0.0333], s_loss:[0.0155] 77 | lr: 2.500000e-05, epoch: [5700/150000], all_loss: [0.0849], m_loss:[0.0575], s_loss:[0.0274] 78 | lr: 2.500000e-05, epoch: [5800/150000], all_loss: [0.0412], m_loss:[0.0295], s_loss:[0.0117] 79 | lr: 2.500000e-05, epoch: [5900/150000], all_loss: [0.0203], m_loss:[0.0110], s_loss:[0.0093] 80 | lr: 2.500000e-05, epoch: [6000/150000], all_loss: [0.0380], m_loss:[0.0251], s_loss:[0.0129] 81 | lr: 2.500000e-05, epoch: [6100/150000], all_loss: [0.1100], m_loss:[0.0697], s_loss:[0.0403] 82 | lr: 2.500000e-05, epoch: [6200/150000], all_loss: [0.0624], m_loss:[0.0339], s_loss:[0.0284] 83 | lr: 2.500000e-05, epoch: [6300/150000], all_loss: [0.0685], m_loss:[0.0373], s_loss:[0.0313] 84 | lr: 2.500000e-05, epoch: [6400/150000], all_loss: [0.1634], m_loss:[0.1211], s_loss:[0.0423] 85 | lr: 2.500000e-05, epoch: [6500/150000], all_loss: [0.0594], m_loss:[0.0388], s_loss:[0.0207] 86 | lr: 2.500000e-05, epoch: [6600/150000], all_loss: [0.0569], m_loss:[0.0420], s_loss:[0.0150] 87 | lr: 2.500000e-05, epoch: [6700/150000], all_loss: [0.0675], m_loss:[0.0418], s_loss:[0.0257] 88 | lr: 2.500000e-05, epoch: [6800/150000], all_loss: [0.0334], m_loss:[0.0232], s_loss:[0.0101] 89 | lr: 2.500000e-05, epoch: [6900/150000], all_loss: [0.0424], m_loss:[0.0271], s_loss:[0.0153] 90 | lr: 2.500000e-05, epoch: [7000/150000], all_loss: [0.0292], m_loss:[0.0179], s_loss:[0.0113] 91 | lr: 2.500000e-05, epoch: [7100/150000], all_loss: [0.0730], m_loss:[0.0454], s_loss:[0.0276] 92 | lr: 2.500000e-05, epoch: [7200/150000], all_loss: [0.0272], m_loss:[0.0173], s_loss:[0.0099] 93 | lr: 2.500000e-05, epoch: [7300/150000], all_loss: [0.0724], m_loss:[0.0489], s_loss:[0.0234] 94 | lr: 2.500000e-05, epoch: [7400/150000], all_loss: [0.0614], m_loss:[0.0364], s_loss:[0.0251] 95 | lr: 2.500000e-05, epoch: [7500/150000], all_loss: [0.0431], m_loss:[0.0304], s_loss:[0.0127] 96 | iCoseg Cosal2015 f: [0.8414], m: [0.0793], f: [0.8424], m: [0.0929] 97 | lr: 2.500000e-05, epoch: [7600/150000], all_loss: [0.0361], m_loss:[0.0215], s_loss:[0.0146] 98 | lr: 2.500000e-05, epoch: [7700/150000], all_loss: [0.0915], m_loss:[0.0617], s_loss:[0.0297] 99 | lr: 2.500000e-05, epoch: [7800/150000], all_loss: [0.0342], m_loss:[0.0229], s_loss:[0.0113] 100 | lr: 2.500000e-05, epoch: [7900/150000], all_loss: [0.0629], m_loss:[0.0397], s_loss:[0.0232] 101 | lr: 2.500000e-05, epoch: [8000/150000], all_loss: [0.0421], m_loss:[0.0273], s_loss:[0.0147] 102 | lr: 2.500000e-05, epoch: [8100/150000], all_loss: [0.0464], m_loss:[0.0312], s_loss:[0.0152] 103 | lr: 2.500000e-05, epoch: [8200/150000], all_loss: [0.0797], m_loss:[0.0562], s_loss:[0.0234] 104 | lr: 2.500000e-05, epoch: [8300/150000], all_loss: [0.0444], m_loss:[0.0255], s_loss:[0.0189] 105 | lr: 2.500000e-05, epoch: [8400/150000], all_loss: [0.0463], m_loss:[0.0242], s_loss:[0.0221] 106 | lr: 2.500000e-05, epoch: [8500/150000], all_loss: [0.0320], m_loss:[0.0194], s_loss:[0.0127] 107 | lr: 2.500000e-05, epoch: [8600/150000], all_loss: [0.0184], m_loss:[0.0115], s_loss:[0.0069] 108 | lr: 2.500000e-05, epoch: [8700/150000], all_loss: [0.0298], m_loss:[0.0216], s_loss:[0.0083] 109 | lr: 2.500000e-05, epoch: [8800/150000], all_loss: [0.0427], m_loss:[0.0286], s_loss:[0.0141] 110 | lr: 2.500000e-05, epoch: [8900/150000], all_loss: [0.0557], m_loss:[0.0345], s_loss:[0.0212] 111 | lr: 2.500000e-05, epoch: [9000/150000], all_loss: [0.0436], m_loss:[0.0310], s_loss:[0.0126] 112 | lr: 2.500000e-05, epoch: [9100/150000], all_loss: [0.0666], m_loss:[0.0393], s_loss:[0.0272] 113 | lr: 2.500000e-05, epoch: [9200/150000], all_loss: [0.0500], m_loss:[0.0372], s_loss:[0.0127] 114 | lr: 2.500000e-05, epoch: [9300/150000], all_loss: [0.0781], m_loss:[0.0533], s_loss:[0.0247] 115 | lr: 2.500000e-05, epoch: [9400/150000], all_loss: [0.0251], m_loss:[0.0131], s_loss:[0.0120] 116 | lr: 2.500000e-05, epoch: [9500/150000], all_loss: [0.0770], m_loss:[0.0535], s_loss:[0.0235] 117 | lr: 2.500000e-05, epoch: [9600/150000], all_loss: [0.0485], m_loss:[0.0323], s_loss:[0.0163] 118 | lr: 2.500000e-05, epoch: [9700/150000], all_loss: [0.0930], m_loss:[0.0661], s_loss:[0.0268] 119 | lr: 2.500000e-05, epoch: [9800/150000], all_loss: [0.0795], m_loss:[0.0577], s_loss:[0.0218] 120 | lr: 2.500000e-05, epoch: [9900/150000], all_loss: [0.0663], m_loss:[0.0449], s_loss:[0.0214] 121 | lr: 2.500000e-05, epoch: [10000/150000], all_loss: [0.0309], m_loss:[0.0203], s_loss:[0.0105] 122 | iCoseg Cosal2015 f: [0.8413], m: [0.0764], f: [0.8433], m: [0.0878] 123 | lr: 2.500000e-05, epoch: [10100/150000], all_loss: [0.0733], m_loss:[0.0490], s_loss:[0.0243] 124 | lr: 2.500000e-05, epoch: [10200/150000], all_loss: [0.0759], m_loss:[0.0485], s_loss:[0.0274] 125 | lr: 2.500000e-05, epoch: [10300/150000], all_loss: [0.0595], m_loss:[0.0400], s_loss:[0.0195] 126 | lr: 2.500000e-05, epoch: [10400/150000], all_loss: [0.0739], m_loss:[0.0516], s_loss:[0.0224] 127 | lr: 2.500000e-05, epoch: [10500/150000], all_loss: [0.0724], m_loss:[0.0444], s_loss:[0.0280] 128 | lr: 2.500000e-05, epoch: [10600/150000], all_loss: [0.0465], m_loss:[0.0307], s_loss:[0.0158] 129 | lr: 2.500000e-05, epoch: [10700/150000], all_loss: [0.0384], m_loss:[0.0255], s_loss:[0.0130] 130 | lr: 2.500000e-05, epoch: [10800/150000], all_loss: [0.0505], m_loss:[0.0269], s_loss:[0.0236] 131 | lr: 2.500000e-05, epoch: [10900/150000], all_loss: [0.0659], m_loss:[0.0464], s_loss:[0.0195] 132 | lr: 2.500000e-05, epoch: [11000/150000], all_loss: [0.0315], m_loss:[0.0189], s_loss:[0.0126] 133 | lr: 2.500000e-05, epoch: [11100/150000], all_loss: [0.0402], m_loss:[0.0219], s_loss:[0.0183] 134 | lr: 2.500000e-05, epoch: [11200/150000], all_loss: [0.0657], m_loss:[0.0449], s_loss:[0.0208] 135 | lr: 2.500000e-05, epoch: [11300/150000], all_loss: [0.0492], m_loss:[0.0325], s_loss:[0.0168] 136 | lr: 2.500000e-05, epoch: [11400/150000], all_loss: [0.0607], m_loss:[0.0337], s_loss:[0.0270] 137 | lr: 2.500000e-05, epoch: [11500/150000], all_loss: [0.0156], m_loss:[0.0087], s_loss:[0.0070] 138 | lr: 2.500000e-05, epoch: [11600/150000], all_loss: [0.0973], m_loss:[0.0636], s_loss:[0.0337] 139 | lr: 2.500000e-05, epoch: [11700/150000], all_loss: [0.0559], m_loss:[0.0397], s_loss:[0.0162] 140 | lr: 2.500000e-05, epoch: [11800/150000], all_loss: [0.0595], m_loss:[0.0384], s_loss:[0.0211] 141 | lr: 2.500000e-05, epoch: [11900/150000], all_loss: [0.0341], m_loss:[0.0241], s_loss:[0.0100] 142 | lr: 2.500000e-05, epoch: [12000/150000], all_loss: [0.0671], m_loss:[0.0494], s_loss:[0.0177] 143 | lr: 2.500000e-05, epoch: [12100/150000], all_loss: [0.0237], m_loss:[0.0149], s_loss:[0.0088] 144 | lr: 2.500000e-05, epoch: [12200/150000], all_loss: [0.0762], m_loss:[0.0512], s_loss:[0.0251] 145 | lr: 2.500000e-05, epoch: [12300/150000], all_loss: [0.0724], m_loss:[0.0547], s_loss:[0.0178] 146 | lr: 2.500000e-05, epoch: [12400/150000], all_loss: [0.0539], m_loss:[0.0372], s_loss:[0.0166] 147 | lr: 2.500000e-05, epoch: [12500/150000], all_loss: [0.0279], m_loss:[0.0192], s_loss:[0.0088] 148 | iCoseg Cosal2015 f: [0.8421], m: [0.0757], f: [0.8461], m: [0.0860] 149 | lr: 2.500000e-05, epoch: [12600/150000], all_loss: [0.0619], m_loss:[0.0371], s_loss:[0.0247] 150 | lr: 2.500000e-05, epoch: [12700/150000], all_loss: [0.0430], m_loss:[0.0244], s_loss:[0.0186] 151 | lr: 2.500000e-05, epoch: [12800/150000], all_loss: [0.0254], m_loss:[0.0151], s_loss:[0.0103] 152 | lr: 2.500000e-05, epoch: [12900/150000], all_loss: [0.0496], m_loss:[0.0328], s_loss:[0.0168] 153 | lr: 2.500000e-05, epoch: [13000/150000], all_loss: [0.0556], m_loss:[0.0398], s_loss:[0.0157] 154 | lr: 2.500000e-05, epoch: [13100/150000], all_loss: [0.0482], m_loss:[0.0327], s_loss:[0.0155] 155 | lr: 2.500000e-05, epoch: [13200/150000], all_loss: [0.0482], m_loss:[0.0352], s_loss:[0.0130] 156 | lr: 2.500000e-05, epoch: [13300/150000], all_loss: [0.0401], m_loss:[0.0278], s_loss:[0.0123] 157 | lr: 2.500000e-05, epoch: [13400/150000], all_loss: [0.0254], m_loss:[0.0163], s_loss:[0.0091] 158 | lr: 2.500000e-05, epoch: [13500/150000], all_loss: [0.0485], m_loss:[0.0273], s_loss:[0.0211] 159 | lr: 2.500000e-05, epoch: [13600/150000], all_loss: [0.0394], m_loss:[0.0227], s_loss:[0.0167] 160 | lr: 2.500000e-05, epoch: [13700/150000], all_loss: [0.0374], m_loss:[0.0252], s_loss:[0.0122] 161 | lr: 2.500000e-05, epoch: [13800/150000], all_loss: [0.0490], m_loss:[0.0323], s_loss:[0.0167] 162 | lr: 2.500000e-05, epoch: [13900/150000], all_loss: [0.0358], m_loss:[0.0216], s_loss:[0.0143] 163 | lr: 2.500000e-05, epoch: [14000/150000], all_loss: [0.0514], m_loss:[0.0354], s_loss:[0.0160] 164 | lr: 2.500000e-05, epoch: [14100/150000], all_loss: [0.0312], m_loss:[0.0166], s_loss:[0.0145] 165 | lr: 2.500000e-05, epoch: [14200/150000], all_loss: [0.1405], m_loss:[0.0893], s_loss:[0.0512] 166 | lr: 2.500000e-05, epoch: [14300/150000], all_loss: [0.0597], m_loss:[0.0374], s_loss:[0.0223] 167 | lr: 2.500000e-05, epoch: [14400/150000], all_loss: [0.0629], m_loss:[0.0384], s_loss:[0.0245] 168 | lr: 2.500000e-05, epoch: [14500/150000], all_loss: [0.0918], m_loss:[0.0555], s_loss:[0.0362] 169 | lr: 2.500000e-05, epoch: [14600/150000], all_loss: [0.0623], m_loss:[0.0417], s_loss:[0.0206] 170 | lr: 2.500000e-05, epoch: [14700/150000], all_loss: [0.0560], m_loss:[0.0417], s_loss:[0.0143] 171 | lr: 2.500000e-05, epoch: [14800/150000], all_loss: [0.0371], m_loss:[0.0248], s_loss:[0.0123] 172 | lr: 2.500000e-05, epoch: [14900/150000], all_loss: [0.0895], m_loss:[0.0616], s_loss:[0.0279] 173 | lr: 2.500000e-05, epoch: [15000/150000], all_loss: [0.0431], m_loss:[0.0252], s_loss:[0.0179] 174 | iCoseg Cosal2015 f: [0.8412], m: [0.0762], f: [0.8484], m: [0.0841] 175 | lr: 2.500000e-05, epoch: [15100/150000], all_loss: [0.0500], m_loss:[0.0275], s_loss:[0.0226] 176 | lr: 2.500000e-05, epoch: [15200/150000], all_loss: [0.1515], m_loss:[0.0910], s_loss:[0.0605] 177 | lr: 2.500000e-05, epoch: [15300/150000], all_loss: [0.0368], m_loss:[0.0225], s_loss:[0.0143] 178 | lr: 2.500000e-05, epoch: [15400/150000], all_loss: [0.0270], m_loss:[0.0180], s_loss:[0.0090] 179 | lr: 2.500000e-05, epoch: [15500/150000], all_loss: [0.0773], m_loss:[0.0447], s_loss:[0.0326] 180 | lr: 2.500000e-05, epoch: [15600/150000], all_loss: [0.0447], m_loss:[0.0331], s_loss:[0.0117] 181 | lr: 2.500000e-05, epoch: [15700/150000], all_loss: [0.0380], m_loss:[0.0181], s_loss:[0.0199] 182 | lr: 2.500000e-05, epoch: [15800/150000], all_loss: [0.0934], m_loss:[0.0634], s_loss:[0.0300] 183 | lr: 2.500000e-05, epoch: [15900/150000], all_loss: [0.0944], m_loss:[0.0630], s_loss:[0.0314] 184 | lr: 2.500000e-05, epoch: [16000/150000], all_loss: [0.0697], m_loss:[0.0480], s_loss:[0.0217] 185 | lr: 2.500000e-05, epoch: [16100/150000], all_loss: [0.0333], m_loss:[0.0207], s_loss:[0.0126] 186 | lr: 2.500000e-05, epoch: [16200/150000], all_loss: [0.0547], m_loss:[0.0348], s_loss:[0.0199] 187 | lr: 2.500000e-05, epoch: [16300/150000], all_loss: [0.0554], m_loss:[0.0362], s_loss:[0.0192] 188 | lr: 2.500000e-05, epoch: [16400/150000], all_loss: [0.0438], m_loss:[0.0272], s_loss:[0.0167] 189 | lr: 2.500000e-05, epoch: [16500/150000], all_loss: [0.0837], m_loss:[0.0549], s_loss:[0.0289] 190 | lr: 2.500000e-05, epoch: [16600/150000], all_loss: [0.0230], m_loss:[0.0129], s_loss:[0.0101] 191 | lr: 2.500000e-05, epoch: [16700/150000], all_loss: [0.0230], m_loss:[0.0131], s_loss:[0.0099] 192 | lr: 2.500000e-05, epoch: [16800/150000], all_loss: [0.0634], m_loss:[0.0422], s_loss:[0.0212] 193 | lr: 2.500000e-05, epoch: [16900/150000], all_loss: [0.1191], m_loss:[0.0698], s_loss:[0.0493] 194 | lr: 2.500000e-05, epoch: [17000/150000], all_loss: [0.0295], m_loss:[0.0170], s_loss:[0.0125] 195 | lr: 2.500000e-05, epoch: [17100/150000], all_loss: [0.0309], m_loss:[0.0199], s_loss:[0.0110] 196 | lr: 2.500000e-05, epoch: [17200/150000], all_loss: [0.0557], m_loss:[0.0349], s_loss:[0.0207] 197 | lr: 2.500000e-05, epoch: [17300/150000], all_loss: [0.0816], m_loss:[0.0618], s_loss:[0.0199] 198 | lr: 2.500000e-05, epoch: [17400/150000], all_loss: [0.0240], m_loss:[0.0144], s_loss:[0.0097] 199 | lr: 2.500000e-05, epoch: [17500/150000], all_loss: [0.0601], m_loss:[0.0373], s_loss:[0.0229] 200 | iCoseg Cosal2015 f: [0.8481], m: [0.0744], f: [0.8512], m: [0.0847] 201 | lr: 2.500000e-05, epoch: [17600/150000], all_loss: [0.0814], m_loss:[0.0434], s_loss:[0.0379] 202 | lr: 2.500000e-05, epoch: [17700/150000], all_loss: [0.2967], m_loss:[0.2219], s_loss:[0.0749] 203 | lr: 2.500000e-05, epoch: [17800/150000], all_loss: [0.4211], m_loss:[0.3101], s_loss:[0.1110] 204 | lr: 2.500000e-05, epoch: [17900/150000], all_loss: [0.0564], m_loss:[0.0415], s_loss:[0.0149] 205 | lr: 2.500000e-05, epoch: [18000/150000], all_loss: [0.0352], m_loss:[0.0251], s_loss:[0.0101] 206 | lr: 2.500000e-05, epoch: [18100/150000], all_loss: [0.1882], m_loss:[0.1203], s_loss:[0.0679] 207 | lr: 2.500000e-05, epoch: [18200/150000], all_loss: [0.0555], m_loss:[0.0378], s_loss:[0.0177] 208 | lr: 2.500000e-05, epoch: [18300/150000], all_loss: [0.0556], m_loss:[0.0326], s_loss:[0.0229] 209 | lr: 2.500000e-05, epoch: [18400/150000], all_loss: [0.0305], m_loss:[0.0170], s_loss:[0.0135] 210 | lr: 2.500000e-05, epoch: [18500/150000], all_loss: [0.0413], m_loss:[0.0269], s_loss:[0.0145] 211 | lr: 2.500000e-05, epoch: [18600/150000], all_loss: [0.0511], m_loss:[0.0365], s_loss:[0.0146] 212 | lr: 2.500000e-05, epoch: [18700/150000], all_loss: [0.0604], m_loss:[0.0416], s_loss:[0.0188] 213 | lr: 2.500000e-05, epoch: [18800/150000], all_loss: [0.0371], m_loss:[0.0262], s_loss:[0.0109] 214 | lr: 2.500000e-05, epoch: [18900/150000], all_loss: [0.0217], m_loss:[0.0117], s_loss:[0.0100] 215 | lr: 2.500000e-05, epoch: [19000/150000], all_loss: [0.0523], m_loss:[0.0396], s_loss:[0.0127] 216 | lr: 2.500000e-05, epoch: [19100/150000], all_loss: [0.0436], m_loss:[0.0247], s_loss:[0.0189] 217 | lr: 2.500000e-05, epoch: [19200/150000], all_loss: [0.0359], m_loss:[0.0259], s_loss:[0.0100] 218 | lr: 2.500000e-05, epoch: [19300/150000], all_loss: [0.1068], m_loss:[0.0746], s_loss:[0.0323] 219 | lr: 2.500000e-05, epoch: [19400/150000], all_loss: [0.0520], m_loss:[0.0301], s_loss:[0.0219] 220 | lr: 2.500000e-05, epoch: [19500/150000], all_loss: [0.0745], m_loss:[0.0490], s_loss:[0.0255] 221 | lr: 2.500000e-05, epoch: [19600/150000], all_loss: [0.0444], m_loss:[0.0320], s_loss:[0.0124] 222 | lr: 2.500000e-05, epoch: [19700/150000], all_loss: [0.0429], m_loss:[0.0281], s_loss:[0.0148] 223 | lr: 2.500000e-05, epoch: [19800/150000], all_loss: [0.0671], m_loss:[0.0424], s_loss:[0.0247] 224 | lr: 2.500000e-05, epoch: [19900/150000], all_loss: [0.0463], m_loss:[0.0300], s_loss:[0.0163] 225 | lr: 2.500000e-05, epoch: [20000/150000], all_loss: [0.0282], m_loss:[0.0162], s_loss:[0.0120] 226 | iCoseg Cosal2015 f: [0.8469], m: [0.0747], f: [0.8466], m: [0.0859] 227 | lr: 2.500000e-05, epoch: [20100/150000], all_loss: [0.0554], m_loss:[0.0398], s_loss:[0.0156] 228 | lr: 2.500000e-05, epoch: [20200/150000], all_loss: [0.0431], m_loss:[0.0249], s_loss:[0.0182] 229 | lr: 2.500000e-05, epoch: [20300/150000], all_loss: [0.0366], m_loss:[0.0206], s_loss:[0.0160] 230 | lr: 2.500000e-05, epoch: [20400/150000], all_loss: [0.0342], m_loss:[0.0218], s_loss:[0.0124] 231 | lr: 2.500000e-05, epoch: [20500/150000], all_loss: [0.0449], m_loss:[0.0325], s_loss:[0.0124] 232 | lr: 2.500000e-05, epoch: [20600/150000], all_loss: [0.0232], m_loss:[0.0158], s_loss:[0.0073] 233 | lr: 2.500000e-05, epoch: [20700/150000], all_loss: [0.0655], m_loss:[0.0438], s_loss:[0.0217] 234 | lr: 2.500000e-05, epoch: [20800/150000], all_loss: [0.0506], m_loss:[0.0334], s_loss:[0.0172] 235 | lr: 2.500000e-05, epoch: [20900/150000], all_loss: [0.1219], m_loss:[0.0877], s_loss:[0.0342] 236 | lr: 2.500000e-05, epoch: [21000/150000], all_loss: [0.1214], m_loss:[0.0787], s_loss:[0.0427] 237 | lr: 2.500000e-05, epoch: [21100/150000], all_loss: [0.0757], m_loss:[0.0492], s_loss:[0.0265] 238 | lr: 2.500000e-05, epoch: [21200/150000], all_loss: [0.0831], m_loss:[0.0587], s_loss:[0.0243] 239 | lr: 2.500000e-05, epoch: [21300/150000], all_loss: [0.0227], m_loss:[0.0137], s_loss:[0.0090] 240 | lr: 2.500000e-05, epoch: [21400/150000], all_loss: [0.0600], m_loss:[0.0374], s_loss:[0.0226] 241 | lr: 2.500000e-05, epoch: [21500/150000], all_loss: [0.0384], m_loss:[0.0251], s_loss:[0.0133] 242 | lr: 2.500000e-05, epoch: [21600/150000], all_loss: [0.0556], m_loss:[0.0354], s_loss:[0.0202] 243 | lr: 2.500000e-05, epoch: [21700/150000], all_loss: [0.0504], m_loss:[0.0322], s_loss:[0.0182] 244 | lr: 2.500000e-05, epoch: [21800/150000], all_loss: [0.1156], m_loss:[0.0842], s_loss:[0.0314] 245 | lr: 2.500000e-05, epoch: [21900/150000], all_loss: [0.0589], m_loss:[0.0390], s_loss:[0.0199] 246 | lr: 2.500000e-05, epoch: [22000/150000], all_loss: [0.0790], m_loss:[0.0489], s_loss:[0.0301] 247 | lr: 2.500000e-05, epoch: [22100/150000], all_loss: [0.0467], m_loss:[0.0327], s_loss:[0.0140] 248 | lr: 2.500000e-05, epoch: [22200/150000], all_loss: [0.0290], m_loss:[0.0184], s_loss:[0.0106] 249 | lr: 2.500000e-05, epoch: [22300/150000], all_loss: [0.0823], m_loss:[0.0615], s_loss:[0.0209] 250 | lr: 2.500000e-05, epoch: [22400/150000], all_loss: [0.0409], m_loss:[0.0317], s_loss:[0.0092] 251 | lr: 2.500000e-05, epoch: [22500/150000], all_loss: [0.1033], m_loss:[0.0690], s_loss:[0.0342] 252 | iCoseg Cosal2015 f: [0.8411], m: [0.0776], f: [0.8466], m: [0.0896] 253 | lr: 2.500000e-05, epoch: [22600/150000], all_loss: [0.0413], m_loss:[0.0284], s_loss:[0.0129] 254 | lr: 2.500000e-05, epoch: [22700/150000], all_loss: [0.0547], m_loss:[0.0374], s_loss:[0.0173] 255 | lr: 2.500000e-05, epoch: [22800/150000], all_loss: [0.0431], m_loss:[0.0301], s_loss:[0.0130] 256 | lr: 2.500000e-05, epoch: [22900/150000], all_loss: [0.0305], m_loss:[0.0196], s_loss:[0.0109] 257 | lr: 2.500000e-05, epoch: [23000/150000], all_loss: [0.0183], m_loss:[0.0113], s_loss:[0.0070] 258 | lr: 2.500000e-05, epoch: [23100/150000], all_loss: [0.0537], m_loss:[0.0380], s_loss:[0.0157] 259 | lr: 2.500000e-05, epoch: [23200/150000], all_loss: [0.0506], m_loss:[0.0353], s_loss:[0.0154] 260 | lr: 2.500000e-05, epoch: [23300/150000], all_loss: [0.0786], m_loss:[0.0525], s_loss:[0.0262] 261 | lr: 2.500000e-05, epoch: [23400/150000], all_loss: [0.0704], m_loss:[0.0493], s_loss:[0.0210] 262 | lr: 2.500000e-05, epoch: [23500/150000], all_loss: [0.1399], m_loss:[0.0923], s_loss:[0.0475] 263 | lr: 2.500000e-05, epoch: [23600/150000], all_loss: [0.0275], m_loss:[0.0161], s_loss:[0.0114] 264 | lr: 2.500000e-05, epoch: [23700/150000], all_loss: [0.2798], m_loss:[0.1860], s_loss:[0.0938] 265 | lr: 2.500000e-05, epoch: [23800/150000], all_loss: [0.0374], m_loss:[0.0265], s_loss:[0.0108] 266 | lr: 2.500000e-05, epoch: [23900/150000], all_loss: [0.0728], m_loss:[0.0461], s_loss:[0.0267] 267 | lr: 2.500000e-05, epoch: [24000/150000], all_loss: [0.0378], m_loss:[0.0268], s_loss:[0.0110] 268 | lr: 2.500000e-05, epoch: [24100/150000], all_loss: [0.1674], m_loss:[0.1069], s_loss:[0.0605] 269 | lr: 2.500000e-05, epoch: [24200/150000], all_loss: [0.0800], m_loss:[0.0586], s_loss:[0.0214] 270 | lr: 2.500000e-05, epoch: [24300/150000], all_loss: [0.0366], m_loss:[0.0236], s_loss:[0.0129] 271 | lr: 2.500000e-05, epoch: [24400/150000], all_loss: [0.1550], m_loss:[0.1014], s_loss:[0.0535] 272 | lr: 2.500000e-05, epoch: [24500/150000], all_loss: [0.0283], m_loss:[0.0173], s_loss:[0.0110] 273 | lr: 2.500000e-05, epoch: [24600/150000], all_loss: [0.0432], m_loss:[0.0262], s_loss:[0.0171] 274 | lr: 2.500000e-05, epoch: [24700/150000], all_loss: [0.0445], m_loss:[0.0269], s_loss:[0.0175] 275 | lr: 2.500000e-05, epoch: [24800/150000], all_loss: [0.0513], m_loss:[0.0334], s_loss:[0.0179] 276 | lr: 2.500000e-05, epoch: [24900/150000], all_loss: [0.1292], m_loss:[0.0915], s_loss:[0.0378] 277 | lr: 2.500000e-05, epoch: [25000/150000], all_loss: [0.0512], m_loss:[0.0374], s_loss:[0.0138] 278 | iCoseg Cosal2015 f: [0.8432], m: [0.0727], f: [0.8462], m: [0.0875] 279 | lr: 1.250000e-05, epoch: [25100/150000], all_loss: [0.0211], m_loss:[0.0122], s_loss:[0.0089] 280 | lr: 1.250000e-05, epoch: [25200/150000], all_loss: [0.0422], m_loss:[0.0293], s_loss:[0.0130] 281 | lr: 1.250000e-05, epoch: [25300/150000], all_loss: [0.0499], m_loss:[0.0271], s_loss:[0.0228] 282 | lr: 1.250000e-05, epoch: [25400/150000], all_loss: [0.0365], m_loss:[0.0218], s_loss:[0.0147] 283 | lr: 1.250000e-05, epoch: [25500/150000], all_loss: [0.1954], m_loss:[0.1351], s_loss:[0.0603] 284 | lr: 1.250000e-05, epoch: [25600/150000], all_loss: [0.0556], m_loss:[0.0415], s_loss:[0.0141] 285 | lr: 1.250000e-05, epoch: [25700/150000], all_loss: [0.0848], m_loss:[0.0607], s_loss:[0.0241] 286 | lr: 1.250000e-05, epoch: [25800/150000], all_loss: [0.0572], m_loss:[0.0354], s_loss:[0.0218] 287 | lr: 1.250000e-05, epoch: [25900/150000], all_loss: [0.0450], m_loss:[0.0250], s_loss:[0.0200] 288 | lr: 1.250000e-05, epoch: [26000/150000], all_loss: [0.1064], m_loss:[0.0705], s_loss:[0.0359] 289 | lr: 1.250000e-05, epoch: [26100/150000], all_loss: [0.0647], m_loss:[0.0467], s_loss:[0.0180] 290 | lr: 1.250000e-05, epoch: [26200/150000], all_loss: [0.0413], m_loss:[0.0310], s_loss:[0.0103] 291 | lr: 1.250000e-05, epoch: [26300/150000], all_loss: [0.0379], m_loss:[0.0258], s_loss:[0.0122] 292 | lr: 1.250000e-05, epoch: [26400/150000], all_loss: [0.0702], m_loss:[0.0456], s_loss:[0.0246] 293 | lr: 1.250000e-05, epoch: [26500/150000], all_loss: [0.0251], m_loss:[0.0151], s_loss:[0.0100] 294 | lr: 1.250000e-05, epoch: [26600/150000], all_loss: [0.0316], m_loss:[0.0201], s_loss:[0.0115] 295 | lr: 1.250000e-05, epoch: [26700/150000], all_loss: [0.0413], m_loss:[0.0267], s_loss:[0.0146] 296 | lr: 1.250000e-05, epoch: [26800/150000], all_loss: [0.0457], m_loss:[0.0326], s_loss:[0.0131] 297 | lr: 1.250000e-05, epoch: [26900/150000], all_loss: [0.0696], m_loss:[0.0426], s_loss:[0.0270] 298 | lr: 1.250000e-05, epoch: [27000/150000], all_loss: [0.0335], m_loss:[0.0221], s_loss:[0.0114] 299 | lr: 1.250000e-05, epoch: [27100/150000], all_loss: [0.0279], m_loss:[0.0160], s_loss:[0.0119] 300 | lr: 1.250000e-05, epoch: [27200/150000], all_loss: [0.0483], m_loss:[0.0348], s_loss:[0.0135] 301 | lr: 1.250000e-05, epoch: [27300/150000], all_loss: [0.0306], m_loss:[0.0184], s_loss:[0.0122] 302 | lr: 1.250000e-05, epoch: [27400/150000], all_loss: [0.0426], m_loss:[0.0249], s_loss:[0.0176] 303 | lr: 1.250000e-05, epoch: [27500/150000], all_loss: [0.0500], m_loss:[0.0273], s_loss:[0.0227] 304 | iCoseg Cosal2015 f: [0.8421], m: [0.0742], f: [0.8468], m: [0.0839] 305 | lr: 1.250000e-05, epoch: [27600/150000], all_loss: [0.0280], m_loss:[0.0168], s_loss:[0.0113] 306 | lr: 1.250000e-05, epoch: [27700/150000], all_loss: [0.0408], m_loss:[0.0283], s_loss:[0.0125] 307 | lr: 1.250000e-05, epoch: [27800/150000], all_loss: [0.0456], m_loss:[0.0339], s_loss:[0.0117] 308 | lr: 1.250000e-05, epoch: [27900/150000], all_loss: [0.0301], m_loss:[0.0202], s_loss:[0.0099] 309 | lr: 1.250000e-05, epoch: [28000/150000], all_loss: [0.0954], m_loss:[0.0569], s_loss:[0.0385] 310 | lr: 1.250000e-05, epoch: [28100/150000], all_loss: [0.0265], m_loss:[0.0170], s_loss:[0.0096] 311 | lr: 1.250000e-05, epoch: [28200/150000], all_loss: [0.0702], m_loss:[0.0495], s_loss:[0.0208] 312 | lr: 1.250000e-05, epoch: [28300/150000], all_loss: [0.1700], m_loss:[0.1044], s_loss:[0.0655] 313 | lr: 1.250000e-05, epoch: [28400/150000], all_loss: [0.0531], m_loss:[0.0360], s_loss:[0.0171] 314 | lr: 1.250000e-05, epoch: [28500/150000], all_loss: [0.0477], m_loss:[0.0326], s_loss:[0.0151] 315 | lr: 1.250000e-05, epoch: [28600/150000], all_loss: [0.0411], m_loss:[0.0293], s_loss:[0.0119] 316 | lr: 1.250000e-05, epoch: [28700/150000], all_loss: [0.0604], m_loss:[0.0410], s_loss:[0.0194] 317 | lr: 1.250000e-05, epoch: [28800/150000], all_loss: [0.1008], m_loss:[0.0670], s_loss:[0.0338] 318 | lr: 1.250000e-05, epoch: [28900/150000], all_loss: [0.1052], m_loss:[0.0652], s_loss:[0.0400] 319 | lr: 1.250000e-05, epoch: [29000/150000], all_loss: [0.0387], m_loss:[0.0268], s_loss:[0.0120] 320 | lr: 1.250000e-05, epoch: [29100/150000], all_loss: [0.1151], m_loss:[0.0747], s_loss:[0.0404] 321 | lr: 1.250000e-05, epoch: [29200/150000], all_loss: [0.0238], m_loss:[0.0169], s_loss:[0.0069] 322 | lr: 1.250000e-05, epoch: [29300/150000], all_loss: [0.0252], m_loss:[0.0189], s_loss:[0.0062] 323 | lr: 1.250000e-05, epoch: [29400/150000], all_loss: [0.0848], m_loss:[0.0574], s_loss:[0.0274] 324 | lr: 1.250000e-05, epoch: [29500/150000], all_loss: [0.0556], m_loss:[0.0414], s_loss:[0.0141] 325 | lr: 1.250000e-05, epoch: [29600/150000], all_loss: [0.0458], m_loss:[0.0294], s_loss:[0.0164] 326 | lr: 1.250000e-05, epoch: [29700/150000], all_loss: [0.0492], m_loss:[0.0321], s_loss:[0.0170] 327 | lr: 1.250000e-05, epoch: [29800/150000], all_loss: [0.0512], m_loss:[0.0324], s_loss:[0.0188] 328 | lr: 1.250000e-05, epoch: [29900/150000], all_loss: [0.0469], m_loss:[0.0318], s_loss:[0.0151] 329 | lr: 1.250000e-05, epoch: [30000/150000], all_loss: [0.0580], m_loss:[0.0364], s_loss:[0.0216] 330 | iCoseg Cosal2015 f: [0.8417], m: [0.0762], f: [0.8480], m: [0.0836] 331 | lr: 1.250000e-05, epoch: [30100/150000], all_loss: [0.0533], m_loss:[0.0263], s_loss:[0.0270] 332 | lr: 1.250000e-05, epoch: [30200/150000], all_loss: [0.1768], m_loss:[0.1320], s_loss:[0.0448] 333 | lr: 1.250000e-05, epoch: [30300/150000], all_loss: [0.0300], m_loss:[0.0198], s_loss:[0.0102] 334 | lr: 1.250000e-05, epoch: [30400/150000], all_loss: [0.0392], m_loss:[0.0279], s_loss:[0.0113] 335 | lr: 1.250000e-05, epoch: [30500/150000], all_loss: [0.0713], m_loss:[0.0492], s_loss:[0.0221] 336 | lr: 1.250000e-05, epoch: [30600/150000], all_loss: [0.0802], m_loss:[0.0597], s_loss:[0.0205] 337 | lr: 1.250000e-05, epoch: [30700/150000], all_loss: [0.0376], m_loss:[0.0255], s_loss:[0.0121] 338 | lr: 1.250000e-05, epoch: [30800/150000], all_loss: [0.0480], m_loss:[0.0328], s_loss:[0.0153] 339 | lr: 1.250000e-05, epoch: [30900/150000], all_loss: [0.0565], m_loss:[0.0384], s_loss:[0.0181] 340 | lr: 1.250000e-05, epoch: [31000/150000], all_loss: [0.0805], m_loss:[0.0529], s_loss:[0.0276] 341 | lr: 1.250000e-05, epoch: [31100/150000], all_loss: [0.0514], m_loss:[0.0277], s_loss:[0.0237] 342 | lr: 1.250000e-05, epoch: [31200/150000], all_loss: [0.0302], m_loss:[0.0199], s_loss:[0.0103] 343 | lr: 1.250000e-05, epoch: [31300/150000], all_loss: [0.0463], m_loss:[0.0292], s_loss:[0.0171] 344 | lr: 1.250000e-05, epoch: [31400/150000], all_loss: [0.0480], m_loss:[0.0333], s_loss:[0.0147] 345 | lr: 1.250000e-05, epoch: [31500/150000], all_loss: [0.0296], m_loss:[0.0128], s_loss:[0.0168] 346 | lr: 1.250000e-05, epoch: [31600/150000], all_loss: [0.0329], m_loss:[0.0203], s_loss:[0.0126] 347 | lr: 1.250000e-05, epoch: [31700/150000], all_loss: [0.0670], m_loss:[0.0426], s_loss:[0.0244] 348 | lr: 1.250000e-05, epoch: [31800/150000], all_loss: [0.0440], m_loss:[0.0230], s_loss:[0.0210] 349 | lr: 1.250000e-05, epoch: [31900/150000], all_loss: [0.0408], m_loss:[0.0274], s_loss:[0.0134] 350 | lr: 1.250000e-05, epoch: [32000/150000], all_loss: [0.0875], m_loss:[0.0594], s_loss:[0.0281] 351 | lr: 1.250000e-05, epoch: [32100/150000], all_loss: [0.0488], m_loss:[0.0302], s_loss:[0.0187] 352 | lr: 1.250000e-05, epoch: [32200/150000], all_loss: [0.0293], m_loss:[0.0190], s_loss:[0.0103] 353 | lr: 1.250000e-05, epoch: [32300/150000], all_loss: [0.0464], m_loss:[0.0295], s_loss:[0.0169] 354 | lr: 1.250000e-05, epoch: [32400/150000], all_loss: [0.0253], m_loss:[0.0182], s_loss:[0.0071] 355 | lr: 1.250000e-05, epoch: [32500/150000], all_loss: [0.0475], m_loss:[0.0335], s_loss:[0.0140] 356 | iCoseg Cosal2015 f: [0.8435], m: [0.0776], f: [0.8516], m: [0.0844] 357 | lr: 1.250000e-05, epoch: [32600/150000], all_loss: [0.1085], m_loss:[0.0658], s_loss:[0.0427] 358 | lr: 1.250000e-05, epoch: [32700/150000], all_loss: [0.0421], m_loss:[0.0254], s_loss:[0.0167] 359 | lr: 1.250000e-05, epoch: [32800/150000], all_loss: [0.0785], m_loss:[0.0605], s_loss:[0.0181] 360 | lr: 1.250000e-05, epoch: [32900/150000], all_loss: [0.0694], m_loss:[0.0510], s_loss:[0.0185] 361 | lr: 1.250000e-05, epoch: [33000/150000], all_loss: [0.0609], m_loss:[0.0438], s_loss:[0.0171] 362 | lr: 1.250000e-05, epoch: [33100/150000], all_loss: [0.0760], m_loss:[0.0464], s_loss:[0.0296] 363 | lr: 1.250000e-05, epoch: [33200/150000], all_loss: [0.0455], m_loss:[0.0321], s_loss:[0.0134] 364 | lr: 1.250000e-05, epoch: [33300/150000], all_loss: [0.0697], m_loss:[0.0491], s_loss:[0.0206] 365 | lr: 1.250000e-05, epoch: [33400/150000], all_loss: [0.0956], m_loss:[0.0687], s_loss:[0.0269] 366 | lr: 1.250000e-05, epoch: [33500/150000], all_loss: [0.0800], m_loss:[0.0573], s_loss:[0.0227] 367 | lr: 1.250000e-05, epoch: [33600/150000], all_loss: [0.0805], m_loss:[0.0481], s_loss:[0.0324] 368 | lr: 1.250000e-05, epoch: [33700/150000], all_loss: [0.0569], m_loss:[0.0418], s_loss:[0.0152] 369 | lr: 1.250000e-05, epoch: [33800/150000], all_loss: [0.0320], m_loss:[0.0163], s_loss:[0.0158] 370 | lr: 1.250000e-05, epoch: [33900/150000], all_loss: [0.0289], m_loss:[0.0189], s_loss:[0.0099] 371 | lr: 1.250000e-05, epoch: [34000/150000], all_loss: [0.0585], m_loss:[0.0387], s_loss:[0.0198] 372 | lr: 1.250000e-05, epoch: [34100/150000], all_loss: [0.0327], m_loss:[0.0223], s_loss:[0.0104] 373 | lr: 1.250000e-05, epoch: [34200/150000], all_loss: [0.1212], m_loss:[0.0803], s_loss:[0.0409] 374 | lr: 1.250000e-05, epoch: [34300/150000], all_loss: [0.0175], m_loss:[0.0109], s_loss:[0.0066] 375 | lr: 1.250000e-05, epoch: [34400/150000], all_loss: [0.0491], m_loss:[0.0361], s_loss:[0.0130] 376 | lr: 1.250000e-05, epoch: [34500/150000], all_loss: [0.0982], m_loss:[0.0589], s_loss:[0.0392] 377 | lr: 1.250000e-05, epoch: [34600/150000], all_loss: [0.0767], m_loss:[0.0504], s_loss:[0.0262] 378 | lr: 1.250000e-05, epoch: [34700/150000], all_loss: [0.0272], m_loss:[0.0156], s_loss:[0.0116] 379 | lr: 1.250000e-05, epoch: [34800/150000], all_loss: [0.0422], m_loss:[0.0242], s_loss:[0.0180] 380 | lr: 1.250000e-05, epoch: [34900/150000], all_loss: [0.0257], m_loss:[0.0173], s_loss:[0.0084] 381 | lr: 1.250000e-05, epoch: [35000/150000], all_loss: [0.0249], m_loss:[0.0169], s_loss:[0.0080] 382 | iCoseg Cosal2015 f: [0.8428], m: [0.0743], f: [0.8503], m: [0.0860] 383 | lr: 1.250000e-05, epoch: [35100/150000], all_loss: [0.0548], m_loss:[0.0346], s_loss:[0.0202] 384 | lr: 1.250000e-05, epoch: [35200/150000], all_loss: [0.0442], m_loss:[0.0313], s_loss:[0.0129] 385 | lr: 1.250000e-05, epoch: [35300/150000], all_loss: [0.0542], m_loss:[0.0343], s_loss:[0.0199] 386 | lr: 1.250000e-05, epoch: [35400/150000], all_loss: [0.0996], m_loss:[0.0725], s_loss:[0.0272] 387 | lr: 1.250000e-05, epoch: [35500/150000], all_loss: [0.0470], m_loss:[0.0300], s_loss:[0.0170] 388 | lr: 1.250000e-05, epoch: [35600/150000], all_loss: [0.0432], m_loss:[0.0294], s_loss:[0.0138] 389 | lr: 1.250000e-05, epoch: [35700/150000], all_loss: [0.0532], m_loss:[0.0371], s_loss:[0.0161] 390 | lr: 1.250000e-05, epoch: [35800/150000], all_loss: [0.0631], m_loss:[0.0427], s_loss:[0.0204] 391 | lr: 1.250000e-05, epoch: [35900/150000], all_loss: [0.0517], m_loss:[0.0377], s_loss:[0.0139] 392 | lr: 1.250000e-05, epoch: [36000/150000], all_loss: [0.0530], m_loss:[0.0406], s_loss:[0.0124] 393 | lr: 1.250000e-05, epoch: [36100/150000], all_loss: [0.0233], m_loss:[0.0150], s_loss:[0.0083] 394 | lr: 1.250000e-05, epoch: [36200/150000], all_loss: [0.2381], m_loss:[0.1604], s_loss:[0.0777] 395 | lr: 1.250000e-05, epoch: [36300/150000], all_loss: [0.0386], m_loss:[0.0276], s_loss:[0.0110] 396 | lr: 1.250000e-05, epoch: [36400/150000], all_loss: [0.0749], m_loss:[0.0412], s_loss:[0.0337] 397 | lr: 1.250000e-05, epoch: [36500/150000], all_loss: [0.0761], m_loss:[0.0533], s_loss:[0.0228] 398 | lr: 1.250000e-05, epoch: [36600/150000], all_loss: [0.0476], m_loss:[0.0287], s_loss:[0.0189] 399 | lr: 1.250000e-05, epoch: [36700/150000], all_loss: [0.0422], m_loss:[0.0298], s_loss:[0.0124] 400 | lr: 1.250000e-05, epoch: [36800/150000], all_loss: [0.0442], m_loss:[0.0308], s_loss:[0.0134] 401 | lr: 1.250000e-05, epoch: [36900/150000], all_loss: [0.0592], m_loss:[0.0414], s_loss:[0.0179] 402 | lr: 1.250000e-05, epoch: [37000/150000], all_loss: [0.0906], m_loss:[0.0716], s_loss:[0.0190] 403 | lr: 1.250000e-05, epoch: [37100/150000], all_loss: [0.0726], m_loss:[0.0494], s_loss:[0.0232] 404 | lr: 1.250000e-05, epoch: [37200/150000], all_loss: [0.0493], m_loss:[0.0319], s_loss:[0.0174] 405 | lr: 1.250000e-05, epoch: [37300/150000], all_loss: [0.0281], m_loss:[0.0182], s_loss:[0.0099] 406 | lr: 1.250000e-05, epoch: [37400/150000], all_loss: [0.0923], m_loss:[0.0670], s_loss:[0.0253] 407 | lr: 1.250000e-05, epoch: [37500/150000], all_loss: [0.0585], m_loss:[0.0384], s_loss:[0.0201] 408 | iCoseg Cosal2015 f: [0.8446], m: [0.0729], f: [0.8518], m: [0.0851] 409 | lr: 1.250000e-05, epoch: [37600/150000], all_loss: [0.1626], m_loss:[0.1109], s_loss:[0.0517] 410 | lr: 1.250000e-05, epoch: [37700/150000], all_loss: [0.0191], m_loss:[0.0116], s_loss:[0.0075] 411 | lr: 1.250000e-05, epoch: [37800/150000], all_loss: [0.0692], m_loss:[0.0427], s_loss:[0.0265] 412 | lr: 1.250000e-05, epoch: [37900/150000], all_loss: [0.0550], m_loss:[0.0335], s_loss:[0.0215] 413 | lr: 1.250000e-05, epoch: [38000/150000], all_loss: [0.0501], m_loss:[0.0354], s_loss:[0.0148] 414 | lr: 1.250000e-05, epoch: [38100/150000], all_loss: [0.1338], m_loss:[0.0868], s_loss:[0.0470] 415 | lr: 1.250000e-05, epoch: [38200/150000], all_loss: [0.0534], m_loss:[0.0287], s_loss:[0.0247] 416 | lr: 1.250000e-05, epoch: [38300/150000], all_loss: [0.0410], m_loss:[0.0268], s_loss:[0.0142] 417 | lr: 1.250000e-05, epoch: [38400/150000], all_loss: [0.0747], m_loss:[0.0535], s_loss:[0.0212] 418 | lr: 1.250000e-05, epoch: [38500/150000], all_loss: [0.0481], m_loss:[0.0300], s_loss:[0.0181] 419 | lr: 1.250000e-05, epoch: [38600/150000], all_loss: [0.0430], m_loss:[0.0281], s_loss:[0.0149] 420 | lr: 1.250000e-05, epoch: [38700/150000], all_loss: [0.0619], m_loss:[0.0398], s_loss:[0.0221] 421 | lr: 1.250000e-05, epoch: [38800/150000], all_loss: [0.0482], m_loss:[0.0329], s_loss:[0.0153] 422 | lr: 1.250000e-05, epoch: [38900/150000], all_loss: [0.0367], m_loss:[0.0234], s_loss:[0.0134] 423 | lr: 1.250000e-05, epoch: [39000/150000], all_loss: [0.0423], m_loss:[0.0279], s_loss:[0.0144] 424 | lr: 1.250000e-05, epoch: [39100/150000], all_loss: [0.0540], m_loss:[0.0306], s_loss:[0.0234] 425 | lr: 1.250000e-05, epoch: [39200/150000], all_loss: [0.1219], m_loss:[0.0883], s_loss:[0.0336] 426 | lr: 1.250000e-05, epoch: [39300/150000], all_loss: [0.0649], m_loss:[0.0420], s_loss:[0.0228] 427 | lr: 1.250000e-05, epoch: [39400/150000], all_loss: [0.0621], m_loss:[0.0424], s_loss:[0.0198] 428 | lr: 1.250000e-05, epoch: [39500/150000], all_loss: [0.0447], m_loss:[0.0326], s_loss:[0.0122] 429 | lr: 1.250000e-05, epoch: [39600/150000], all_loss: [0.0687], m_loss:[0.0521], s_loss:[0.0166] 430 | lr: 1.250000e-05, epoch: [39700/150000], all_loss: [0.1238], m_loss:[0.0908], s_loss:[0.0329] 431 | lr: 1.250000e-05, epoch: [39800/150000], all_loss: [0.0421], m_loss:[0.0289], s_loss:[0.0131] 432 | lr: 1.250000e-05, epoch: [39900/150000], all_loss: [0.0321], m_loss:[0.0188], s_loss:[0.0132] 433 | lr: 1.250000e-05, epoch: [40000/150000], all_loss: [0.0450], m_loss:[0.0286], s_loss:[0.0164] 434 | iCoseg Cosal2015 f: [0.8437], m: [0.0766], f: [0.8499], m: [0.0870] 435 | lr: 1.250000e-05, epoch: [40100/150000], all_loss: [0.0536], m_loss:[0.0309], s_loss:[0.0227] 436 | lr: 1.250000e-05, epoch: [40200/150000], all_loss: [0.0257], m_loss:[0.0160], s_loss:[0.0097] 437 | lr: 1.250000e-05, epoch: [40300/150000], all_loss: [0.0338], m_loss:[0.0195], s_loss:[0.0143] 438 | lr: 1.250000e-05, epoch: [40400/150000], all_loss: [0.0639], m_loss:[0.0356], s_loss:[0.0283] 439 | lr: 1.250000e-05, epoch: [40500/150000], all_loss: [0.0574], m_loss:[0.0354], s_loss:[0.0220] 440 | lr: 1.250000e-05, epoch: [40600/150000], all_loss: [0.0358], m_loss:[0.0231], s_loss:[0.0126] 441 | lr: 1.250000e-05, epoch: [40700/150000], all_loss: [0.0495], m_loss:[0.0362], s_loss:[0.0133] 442 | lr: 1.250000e-05, epoch: [40800/150000], all_loss: [0.0246], m_loss:[0.0155], s_loss:[0.0091] 443 | lr: 1.250000e-05, epoch: [40900/150000], all_loss: [0.0650], m_loss:[0.0440], s_loss:[0.0210] 444 | lr: 1.250000e-05, epoch: [41000/150000], all_loss: [0.1661], m_loss:[0.1144], s_loss:[0.0517] 445 | lr: 1.250000e-05, epoch: [41100/150000], all_loss: [0.0141], m_loss:[0.0094], s_loss:[0.0047] 446 | lr: 1.250000e-05, epoch: [41200/150000], all_loss: [0.0935], m_loss:[0.0598], s_loss:[0.0337] 447 | lr: 1.250000e-05, epoch: [41300/150000], all_loss: [0.0963], m_loss:[0.0637], s_loss:[0.0327] 448 | lr: 1.250000e-05, epoch: [41400/150000], all_loss: [0.0488], m_loss:[0.0301], s_loss:[0.0187] 449 | lr: 1.250000e-05, epoch: [41500/150000], all_loss: [0.0337], m_loss:[0.0198], s_loss:[0.0139] 450 | lr: 1.250000e-05, epoch: [41600/150000], all_loss: [0.0301], m_loss:[0.0193], s_loss:[0.0109] 451 | lr: 1.250000e-05, epoch: [41700/150000], all_loss: [0.0587], m_loss:[0.0430], s_loss:[0.0157] 452 | lr: 1.250000e-05, epoch: [41800/150000], all_loss: [0.0686], m_loss:[0.0458], s_loss:[0.0228] 453 | lr: 1.250000e-05, epoch: [41900/150000], all_loss: [0.0425], m_loss:[0.0288], s_loss:[0.0137] 454 | lr: 1.250000e-05, epoch: [42000/150000], all_loss: [0.0458], m_loss:[0.0266], s_loss:[0.0192] 455 | lr: 1.250000e-05, epoch: [42100/150000], all_loss: [0.0147], m_loss:[0.0095], s_loss:[0.0053] 456 | lr: 1.250000e-05, epoch: [42200/150000], all_loss: [0.0542], m_loss:[0.0299], s_loss:[0.0243] 457 | lr: 1.250000e-05, epoch: [42300/150000], all_loss: [0.1330], m_loss:[0.0985], s_loss:[0.0345] 458 | lr: 1.250000e-05, epoch: [42400/150000], all_loss: [0.1666], m_loss:[0.1177], s_loss:[0.0489] 459 | lr: 1.250000e-05, epoch: [42500/150000], all_loss: [0.0290], m_loss:[0.0205], s_loss:[0.0085] 460 | iCoseg Cosal2015 f: [0.8446], m: [0.0732], f: [0.8528], m: [0.0822] 461 | lr: 1.250000e-05, epoch: [42600/150000], all_loss: [0.0299], m_loss:[0.0195], s_loss:[0.0104] 462 | lr: 1.250000e-05, epoch: [42700/150000], all_loss: [0.0396], m_loss:[0.0230], s_loss:[0.0167] 463 | lr: 1.250000e-05, epoch: [42800/150000], all_loss: [0.0209], m_loss:[0.0125], s_loss:[0.0084] 464 | lr: 1.250000e-05, epoch: [42900/150000], all_loss: [0.0504], m_loss:[0.0356], s_loss:[0.0148] 465 | lr: 1.250000e-05, epoch: [43000/150000], all_loss: [0.0657], m_loss:[0.0409], s_loss:[0.0248] 466 | lr: 1.250000e-05, epoch: [43100/150000], all_loss: [0.0447], m_loss:[0.0304], s_loss:[0.0143] 467 | lr: 1.250000e-05, epoch: [43200/150000], all_loss: [0.0242], m_loss:[0.0145], s_loss:[0.0096] 468 | lr: 1.250000e-05, epoch: [43300/150000], all_loss: [0.0542], m_loss:[0.0412], s_loss:[0.0129] 469 | lr: 1.250000e-05, epoch: [43400/150000], all_loss: [0.2218], m_loss:[0.1636], s_loss:[0.0582] 470 | lr: 1.250000e-05, epoch: [43500/150000], all_loss: [0.0404], m_loss:[0.0287], s_loss:[0.0118] 471 | lr: 1.250000e-05, epoch: [43600/150000], all_loss: [0.1388], m_loss:[0.1017], s_loss:[0.0371] 472 | lr: 1.250000e-05, epoch: [43700/150000], all_loss: [0.0514], m_loss:[0.0371], s_loss:[0.0143] 473 | lr: 1.250000e-05, epoch: [43800/150000], all_loss: [0.0591], m_loss:[0.0378], s_loss:[0.0213] 474 | lr: 1.250000e-05, epoch: [43900/150000], all_loss: [0.0295], m_loss:[0.0189], s_loss:[0.0107] 475 | lr: 1.250000e-05, epoch: [44000/150000], all_loss: [0.0366], m_loss:[0.0220], s_loss:[0.0145] 476 | lr: 1.250000e-05, epoch: [44100/150000], all_loss: [0.0921], m_loss:[0.0545], s_loss:[0.0376] 477 | lr: 1.250000e-05, epoch: [44200/150000], all_loss: [0.1382], m_loss:[0.0945], s_loss:[0.0437] 478 | lr: 1.250000e-05, epoch: [44300/150000], all_loss: [0.0740], m_loss:[0.0549], s_loss:[0.0191] 479 | lr: 1.250000e-05, epoch: [44400/150000], all_loss: [0.0473], m_loss:[0.0348], s_loss:[0.0125] 480 | lr: 1.250000e-05, epoch: [44500/150000], all_loss: [0.0961], m_loss:[0.0586], s_loss:[0.0376] 481 | lr: 1.250000e-05, epoch: [44600/150000], all_loss: [0.0219], m_loss:[0.0148], s_loss:[0.0071] 482 | lr: 1.250000e-05, epoch: [44700/150000], all_loss: [0.0773], m_loss:[0.0454], s_loss:[0.0319] 483 | lr: 1.250000e-05, epoch: [44800/150000], all_loss: [0.0407], m_loss:[0.0271], s_loss:[0.0136] 484 | lr: 1.250000e-05, epoch: [44900/150000], all_loss: [0.0212], m_loss:[0.0130], s_loss:[0.0082] 485 | lr: 1.250000e-05, epoch: [45000/150000], all_loss: [0.0629], m_loss:[0.0511], s_loss:[0.0118] 486 | iCoseg Cosal2015 f: [0.8462], m: [0.0700], f: [0.8547], m: [0.0770] 487 | lr: 1.250000e-05, epoch: [45100/150000], all_loss: [0.0567], m_loss:[0.0340], s_loss:[0.0227] 488 | lr: 1.250000e-05, epoch: [45200/150000], all_loss: [0.0533], m_loss:[0.0310], s_loss:[0.0223] 489 | lr: 1.250000e-05, epoch: [45300/150000], all_loss: [0.0395], m_loss:[0.0204], s_loss:[0.0192] 490 | lr: 1.250000e-05, epoch: [45400/150000], all_loss: [0.0609], m_loss:[0.0423], s_loss:[0.0186] 491 | lr: 1.250000e-05, epoch: [45500/150000], all_loss: [0.0327], m_loss:[0.0216], s_loss:[0.0111] 492 | lr: 1.250000e-05, epoch: [45600/150000], all_loss: [0.0468], m_loss:[0.0326], s_loss:[0.0142] 493 | lr: 1.250000e-05, epoch: [45700/150000], all_loss: [0.0453], m_loss:[0.0310], s_loss:[0.0143] 494 | lr: 1.250000e-05, epoch: [45800/150000], all_loss: [0.0237], m_loss:[0.0159], s_loss:[0.0078] 495 | lr: 1.250000e-05, epoch: [45900/150000], all_loss: [0.0315], m_loss:[0.0216], s_loss:[0.0099] 496 | lr: 1.250000e-05, epoch: [46000/150000], all_loss: [0.0285], m_loss:[0.0193], s_loss:[0.0092] 497 | lr: 1.250000e-05, epoch: [46100/150000], all_loss: [0.0314], m_loss:[0.0224], s_loss:[0.0091] 498 | lr: 1.250000e-05, epoch: [46200/150000], all_loss: [0.0435], m_loss:[0.0246], s_loss:[0.0189] 499 | lr: 1.250000e-05, epoch: [46300/150000], all_loss: [0.0665], m_loss:[0.0463], s_loss:[0.0202] 500 | lr: 1.250000e-05, epoch: [46400/150000], all_loss: [0.0882], m_loss:[0.0581], s_loss:[0.0301] 501 | lr: 1.250000e-05, epoch: [46500/150000], all_loss: [0.0512], m_loss:[0.0395], s_loss:[0.0117] 502 | lr: 1.250000e-05, epoch: [46600/150000], all_loss: [0.0225], m_loss:[0.0112], s_loss:[0.0112] 503 | lr: 1.250000e-05, epoch: [46700/150000], all_loss: [0.1067], m_loss:[0.0640], s_loss:[0.0426] 504 | lr: 1.250000e-05, epoch: [46800/150000], all_loss: [0.0591], m_loss:[0.0384], s_loss:[0.0206] 505 | lr: 1.250000e-05, epoch: [46900/150000], all_loss: [0.0292], m_loss:[0.0168], s_loss:[0.0125] 506 | lr: 1.250000e-05, epoch: [47000/150000], all_loss: [0.0357], m_loss:[0.0194], s_loss:[0.0162] 507 | lr: 1.250000e-05, epoch: [47100/150000], all_loss: [0.1658], m_loss:[0.1100], s_loss:[0.0559] 508 | lr: 1.250000e-05, epoch: [47200/150000], all_loss: [0.0314], m_loss:[0.0219], s_loss:[0.0095] 509 | lr: 1.250000e-05, epoch: [47300/150000], all_loss: [0.0338], m_loss:[0.0226], s_loss:[0.0113] 510 | lr: 1.250000e-05, epoch: [47400/150000], all_loss: [0.0245], m_loss:[0.0153], s_loss:[0.0093] 511 | lr: 1.250000e-05, epoch: [47500/150000], all_loss: [0.0916], m_loss:[0.0629], s_loss:[0.0286] 512 | iCoseg Cosal2015 f: [0.8444], m: [0.0731], f: [0.8506], m: [0.0844] 513 | lr: 1.250000e-05, epoch: [47600/150000], all_loss: [0.0307], m_loss:[0.0160], s_loss:[0.0147] 514 | lr: 1.250000e-05, epoch: [47700/150000], all_loss: [0.0572], m_loss:[0.0447], s_loss:[0.0125] 515 | lr: 1.250000e-05, epoch: [47800/150000], all_loss: [0.0333], m_loss:[0.0226], s_loss:[0.0107] 516 | lr: 1.250000e-05, epoch: [47900/150000], all_loss: [0.0249], m_loss:[0.0139], s_loss:[0.0111] 517 | lr: 1.250000e-05, epoch: [48000/150000], all_loss: [0.1106], m_loss:[0.0830], s_loss:[0.0276] 518 | lr: 1.250000e-05, epoch: [48100/150000], all_loss: [0.0388], m_loss:[0.0216], s_loss:[0.0172] 519 | lr: 1.250000e-05, epoch: [48200/150000], all_loss: [0.0829], m_loss:[0.0556], s_loss:[0.0274] 520 | lr: 1.250000e-05, epoch: [48300/150000], all_loss: [0.0625], m_loss:[0.0462], s_loss:[0.0163] 521 | lr: 1.250000e-05, epoch: [48400/150000], all_loss: [0.0513], m_loss:[0.0311], s_loss:[0.0202] 522 | lr: 1.250000e-05, epoch: [48500/150000], all_loss: [0.0650], m_loss:[0.0426], s_loss:[0.0224] 523 | lr: 1.250000e-05, epoch: [48600/150000], all_loss: [0.0390], m_loss:[0.0274], s_loss:[0.0116] 524 | lr: 1.250000e-05, epoch: [48700/150000], all_loss: [0.0443], m_loss:[0.0308], s_loss:[0.0135] 525 | lr: 1.250000e-05, epoch: [48800/150000], all_loss: [0.0297], m_loss:[0.0181], s_loss:[0.0116] 526 | lr: 1.250000e-05, epoch: [48900/150000], all_loss: [0.0424], m_loss:[0.0267], s_loss:[0.0157] 527 | lr: 1.250000e-05, epoch: [49000/150000], all_loss: [0.0131], m_loss:[0.0068], s_loss:[0.0063] 528 | lr: 1.250000e-05, epoch: [49100/150000], all_loss: [0.1402], m_loss:[0.0983], s_loss:[0.0419] 529 | lr: 1.250000e-05, epoch: [49200/150000], all_loss: [0.0565], m_loss:[0.0426], s_loss:[0.0139] 530 | lr: 1.250000e-05, epoch: [49300/150000], all_loss: [0.0729], m_loss:[0.0457], s_loss:[0.0272] 531 | lr: 1.250000e-05, epoch: [49400/150000], all_loss: [0.0735], m_loss:[0.0401], s_loss:[0.0334] 532 | lr: 1.250000e-05, epoch: [49500/150000], all_loss: [0.0383], m_loss:[0.0267], s_loss:[0.0115] 533 | lr: 1.250000e-05, epoch: [49600/150000], all_loss: [0.3039], m_loss:[0.2129], s_loss:[0.0910] 534 | lr: 1.250000e-05, epoch: [49700/150000], all_loss: [0.0388], m_loss:[0.0269], s_loss:[0.0119] 535 | lr: 1.250000e-05, epoch: [49800/150000], all_loss: [0.0550], m_loss:[0.0370], s_loss:[0.0180] 536 | lr: 1.250000e-05, epoch: [49900/150000], all_loss: [0.0535], m_loss:[0.0341], s_loss:[0.0194] 537 | lr: 1.250000e-05, epoch: [50000/150000], all_loss: [0.0561], m_loss:[0.0387], s_loss:[0.0175] 538 | iCoseg Cosal2015 f: [0.8406], m: [0.0767], f: [0.8546], m: [0.0820] 539 | lr: 6.250000e-06, epoch: [50100/150000], all_loss: [0.0327], m_loss:[0.0205], s_loss:[0.0122] 540 | lr: 6.250000e-06, epoch: [50200/150000], all_loss: [0.0368], m_loss:[0.0252], s_loss:[0.0115] 541 | lr: 6.250000e-06, epoch: [50300/150000], all_loss: [0.0341], m_loss:[0.0242], s_loss:[0.0099] 542 | lr: 6.250000e-06, epoch: [50400/150000], all_loss: [0.0382], m_loss:[0.0207], s_loss:[0.0175] 543 | lr: 6.250000e-06, epoch: [50500/150000], all_loss: [0.0485], m_loss:[0.0309], s_loss:[0.0176] 544 | lr: 6.250000e-06, epoch: [50600/150000], all_loss: [0.0236], m_loss:[0.0144], s_loss:[0.0092] 545 | lr: 6.250000e-06, epoch: [50700/150000], all_loss: [0.0505], m_loss:[0.0307], s_loss:[0.0198] 546 | lr: 6.250000e-06, epoch: [50800/150000], all_loss: [0.0404], m_loss:[0.0263], s_loss:[0.0142] 547 | lr: 6.250000e-06, epoch: [50900/150000], all_loss: [0.0615], m_loss:[0.0433], s_loss:[0.0182] 548 | lr: 6.250000e-06, epoch: [51000/150000], all_loss: [0.0789], m_loss:[0.0533], s_loss:[0.0257] 549 | lr: 6.250000e-06, epoch: [51100/150000], all_loss: [0.0601], m_loss:[0.0348], s_loss:[0.0252] 550 | lr: 6.250000e-06, epoch: [51200/150000], all_loss: [0.0382], m_loss:[0.0260], s_loss:[0.0122] 551 | lr: 6.250000e-06, epoch: [51300/150000], all_loss: [0.1391], m_loss:[0.0825], s_loss:[0.0566] 552 | lr: 6.250000e-06, epoch: [51400/150000], all_loss: [0.0341], m_loss:[0.0238], s_loss:[0.0103] 553 | lr: 6.250000e-06, epoch: [51500/150000], all_loss: [0.0614], m_loss:[0.0423], s_loss:[0.0191] 554 | lr: 6.250000e-06, epoch: [51600/150000], all_loss: [0.0465], m_loss:[0.0297], s_loss:[0.0168] 555 | lr: 6.250000e-06, epoch: [51700/150000], all_loss: [0.0648], m_loss:[0.0486], s_loss:[0.0163] 556 | lr: 6.250000e-06, epoch: [51800/150000], all_loss: [0.0753], m_loss:[0.0461], s_loss:[0.0293] 557 | lr: 6.250000e-06, epoch: [51900/150000], all_loss: [0.0193], m_loss:[0.0132], s_loss:[0.0061] 558 | lr: 6.250000e-06, epoch: [52000/150000], all_loss: [0.0285], m_loss:[0.0148], s_loss:[0.0137] 559 | lr: 6.250000e-06, epoch: [52100/150000], all_loss: [0.0504], m_loss:[0.0348], s_loss:[0.0156] 560 | lr: 6.250000e-06, epoch: [52200/150000], all_loss: [0.1381], m_loss:[0.0881], s_loss:[0.0500] 561 | lr: 6.250000e-06, epoch: [52300/150000], all_loss: [0.0335], m_loss:[0.0200], s_loss:[0.0135] 562 | lr: 6.250000e-06, epoch: [52400/150000], all_loss: [0.0299], m_loss:[0.0190], s_loss:[0.0109] 563 | lr: 6.250000e-06, epoch: [52500/150000], all_loss: [0.0857], m_loss:[0.0605], s_loss:[0.0252] 564 | iCoseg Cosal2015 f: [0.8411], m: [0.0762], f: [0.8535], m: [0.0796] 565 | lr: 6.250000e-06, epoch: [52600/150000], all_loss: [0.0580], m_loss:[0.0430], s_loss:[0.0150] 566 | lr: 6.250000e-06, epoch: [52700/150000], all_loss: [0.0303], m_loss:[0.0210], s_loss:[0.0093] 567 | lr: 6.250000e-06, epoch: [52800/150000], all_loss: [0.0239], m_loss:[0.0133], s_loss:[0.0106] 568 | lr: 6.250000e-06, epoch: [52900/150000], all_loss: [0.0327], m_loss:[0.0176], s_loss:[0.0150] 569 | lr: 6.250000e-06, epoch: [53000/150000], all_loss: [0.0752], m_loss:[0.0524], s_loss:[0.0228] 570 | lr: 6.250000e-06, epoch: [53100/150000], all_loss: [0.0261], m_loss:[0.0180], s_loss:[0.0081] 571 | lr: 6.250000e-06, epoch: [53200/150000], all_loss: [0.0523], m_loss:[0.0334], s_loss:[0.0189] 572 | lr: 6.250000e-06, epoch: [53300/150000], all_loss: [0.0716], m_loss:[0.0410], s_loss:[0.0306] 573 | lr: 6.250000e-06, epoch: [53400/150000], all_loss: [0.0327], m_loss:[0.0189], s_loss:[0.0138] 574 | lr: 6.250000e-06, epoch: [53500/150000], all_loss: [0.0141], m_loss:[0.0084], s_loss:[0.0058] 575 | lr: 6.250000e-06, epoch: [53600/150000], all_loss: [0.0472], m_loss:[0.0330], s_loss:[0.0142] 576 | lr: 6.250000e-06, epoch: [53700/150000], all_loss: [0.0543], m_loss:[0.0320], s_loss:[0.0224] 577 | lr: 6.250000e-06, epoch: [53800/150000], all_loss: [0.0469], m_loss:[0.0291], s_loss:[0.0178] 578 | lr: 6.250000e-06, epoch: [53900/150000], all_loss: [0.0529], m_loss:[0.0346], s_loss:[0.0183] 579 | lr: 6.250000e-06, epoch: [54000/150000], all_loss: [0.0289], m_loss:[0.0183], s_loss:[0.0106] 580 | lr: 6.250000e-06, epoch: [54100/150000], all_loss: [0.0255], m_loss:[0.0176], s_loss:[0.0079] 581 | lr: 6.250000e-06, epoch: [54200/150000], all_loss: [0.0230], m_loss:[0.0149], s_loss:[0.0082] 582 | lr: 6.250000e-06, epoch: [54300/150000], all_loss: [0.0497], m_loss:[0.0331], s_loss:[0.0166] 583 | lr: 6.250000e-06, epoch: [54400/150000], all_loss: [0.0293], m_loss:[0.0185], s_loss:[0.0108] 584 | lr: 6.250000e-06, epoch: [54500/150000], all_loss: [0.0440], m_loss:[0.0303], s_loss:[0.0137] 585 | lr: 6.250000e-06, epoch: [54600/150000], all_loss: [0.0425], m_loss:[0.0276], s_loss:[0.0149] 586 | lr: 6.250000e-06, epoch: [54700/150000], all_loss: [0.0498], m_loss:[0.0373], s_loss:[0.0125] 587 | lr: 6.250000e-06, epoch: [54800/150000], all_loss: [0.0477], m_loss:[0.0339], s_loss:[0.0138] 588 | lr: 6.250000e-06, epoch: [54900/150000], all_loss: [0.0446], m_loss:[0.0265], s_loss:[0.0182] 589 | lr: 6.250000e-06, epoch: [55000/150000], all_loss: [0.2591], m_loss:[0.1744], s_loss:[0.0848] 590 | iCoseg Cosal2015 f: [0.8413], m: [0.0744], f: [0.8531], m: [0.0798] 591 | lr: 6.250000e-06, epoch: [55100/150000], all_loss: [0.0524], m_loss:[0.0404], s_loss:[0.0120] 592 | lr: 6.250000e-06, epoch: [55200/150000], all_loss: [0.0418], m_loss:[0.0290], s_loss:[0.0128] 593 | lr: 6.250000e-06, epoch: [55300/150000], all_loss: [0.0537], m_loss:[0.0334], s_loss:[0.0203] 594 | lr: 6.250000e-06, epoch: [55400/150000], all_loss: [0.0566], m_loss:[0.0426], s_loss:[0.0140] 595 | lr: 6.250000e-06, epoch: [55500/150000], all_loss: [0.0803], m_loss:[0.0542], s_loss:[0.0261] 596 | lr: 6.250000e-06, epoch: [55600/150000], all_loss: [0.0421], m_loss:[0.0254], s_loss:[0.0167] 597 | lr: 6.250000e-06, epoch: [55700/150000], all_loss: [0.0614], m_loss:[0.0317], s_loss:[0.0297] 598 | lr: 6.250000e-06, epoch: [55800/150000], all_loss: [0.0339], m_loss:[0.0223], s_loss:[0.0116] 599 | lr: 6.250000e-06, epoch: [55900/150000], all_loss: [0.0371], m_loss:[0.0268], s_loss:[0.0103] 600 | lr: 6.250000e-06, epoch: [56000/150000], all_loss: [0.0695], m_loss:[0.0491], s_loss:[0.0204] 601 | lr: 6.250000e-06, epoch: [56100/150000], all_loss: [0.0231], m_loss:[0.0141], s_loss:[0.0090] 602 | lr: 6.250000e-06, epoch: [56200/150000], all_loss: [0.0367], m_loss:[0.0264], s_loss:[0.0103] 603 | lr: 6.250000e-06, epoch: [56300/150000], all_loss: [0.0481], m_loss:[0.0356], s_loss:[0.0125] 604 | lr: 6.250000e-06, epoch: [56400/150000], all_loss: [0.0459], m_loss:[0.0303], s_loss:[0.0156] 605 | lr: 6.250000e-06, epoch: [56500/150000], all_loss: [0.0436], m_loss:[0.0327], s_loss:[0.0109] 606 | lr: 6.250000e-06, epoch: [56600/150000], all_loss: [0.0520], m_loss:[0.0399], s_loss:[0.0122] 607 | lr: 6.250000e-06, epoch: [56700/150000], all_loss: [0.0441], m_loss:[0.0278], s_loss:[0.0163] 608 | lr: 6.250000e-06, epoch: [56800/150000], all_loss: [0.0563], m_loss:[0.0397], s_loss:[0.0166] 609 | lr: 6.250000e-06, epoch: [56900/150000], all_loss: [0.0694], m_loss:[0.0354], s_loss:[0.0340] 610 | lr: 6.250000e-06, epoch: [57000/150000], all_loss: [0.0702], m_loss:[0.0458], s_loss:[0.0243] 611 | lr: 6.250000e-06, epoch: [57100/150000], all_loss: [0.0279], m_loss:[0.0166], s_loss:[0.0113] 612 | lr: 6.250000e-06, epoch: [57200/150000], all_loss: [0.0588], m_loss:[0.0384], s_loss:[0.0204] 613 | lr: 6.250000e-06, epoch: [57300/150000], all_loss: [0.0592], m_loss:[0.0432], s_loss:[0.0160] 614 | lr: 6.250000e-06, epoch: [57400/150000], all_loss: [0.0423], m_loss:[0.0295], s_loss:[0.0129] 615 | lr: 6.250000e-06, epoch: [57500/150000], all_loss: [0.1454], m_loss:[0.0979], s_loss:[0.0476] 616 | iCoseg Cosal2015 f: [0.8424], m: [0.0727], f: [0.8522], m: [0.0801] 617 | lr: 6.250000e-06, epoch: [57600/150000], all_loss: [0.0706], m_loss:[0.0439], s_loss:[0.0267] 618 | lr: 6.250000e-06, epoch: [57700/150000], all_loss: [0.0579], m_loss:[0.0352], s_loss:[0.0227] 619 | lr: 6.250000e-06, epoch: [57800/150000], all_loss: [0.0449], m_loss:[0.0285], s_loss:[0.0164] 620 | lr: 6.250000e-06, epoch: [57900/150000], all_loss: [0.0214], m_loss:[0.0131], s_loss:[0.0083] 621 | lr: 6.250000e-06, epoch: [58000/150000], all_loss: [0.0632], m_loss:[0.0480], s_loss:[0.0152] 622 | lr: 6.250000e-06, epoch: [58100/150000], all_loss: [0.0171], m_loss:[0.0097], s_loss:[0.0075] 623 | lr: 6.250000e-06, epoch: [58200/150000], all_loss: [0.0464], m_loss:[0.0301], s_loss:[0.0163] 624 | lr: 6.250000e-06, epoch: [58300/150000], all_loss: [0.0679], m_loss:[0.0476], s_loss:[0.0203] 625 | lr: 6.250000e-06, epoch: [58400/150000], all_loss: [0.0388], m_loss:[0.0300], s_loss:[0.0088] 626 | lr: 6.250000e-06, epoch: [58500/150000], all_loss: [0.1197], m_loss:[0.0829], s_loss:[0.0367] 627 | lr: 6.250000e-06, epoch: [58600/150000], all_loss: [0.0523], m_loss:[0.0375], s_loss:[0.0148] 628 | lr: 6.250000e-06, epoch: [58700/150000], all_loss: [0.1055], m_loss:[0.0743], s_loss:[0.0312] 629 | lr: 6.250000e-06, epoch: [58800/150000], all_loss: [0.0355], m_loss:[0.0234], s_loss:[0.0122] 630 | lr: 6.250000e-06, epoch: [58900/150000], all_loss: [0.0387], m_loss:[0.0266], s_loss:[0.0121] 631 | lr: 6.250000e-06, epoch: [59000/150000], all_loss: [0.0437], m_loss:[0.0285], s_loss:[0.0152] 632 | lr: 6.250000e-06, epoch: [59100/150000], all_loss: [0.0294], m_loss:[0.0214], s_loss:[0.0080] 633 | lr: 6.250000e-06, epoch: [59200/150000], all_loss: [0.0526], m_loss:[0.0388], s_loss:[0.0138] 634 | lr: 6.250000e-06, epoch: [59300/150000], all_loss: [0.0375], m_loss:[0.0250], s_loss:[0.0125] 635 | lr: 6.250000e-06, epoch: [59400/150000], all_loss: [0.1113], m_loss:[0.0865], s_loss:[0.0248] 636 | lr: 6.250000e-06, epoch: [59500/150000], all_loss: [0.0281], m_loss:[0.0164], s_loss:[0.0117] 637 | lr: 6.250000e-06, epoch: [59600/150000], all_loss: [0.0684], m_loss:[0.0478], s_loss:[0.0206] 638 | lr: 6.250000e-06, epoch: [59700/150000], all_loss: [0.0493], m_loss:[0.0346], s_loss:[0.0146] 639 | lr: 6.250000e-06, epoch: [59800/150000], all_loss: [0.0189], m_loss:[0.0091], s_loss:[0.0098] 640 | lr: 6.250000e-06, epoch: [59900/150000], all_loss: [0.0772], m_loss:[0.0534], s_loss:[0.0238] 641 | lr: 6.250000e-06, epoch: [60000/150000], all_loss: [0.1138], m_loss:[0.0745], s_loss:[0.0392] 642 | iCoseg Cosal2015 f: [0.8438], m: [0.0721], f: [0.8529], m: [0.0826] 643 | lr: 6.250000e-06, epoch: [60100/150000], all_loss: [0.0411], m_loss:[0.0269], s_loss:[0.0142] 644 | lr: 6.250000e-06, epoch: [60200/150000], all_loss: [0.1066], m_loss:[0.0849], s_loss:[0.0217] 645 | lr: 6.250000e-06, epoch: [60300/150000], all_loss: [0.0519], m_loss:[0.0351], s_loss:[0.0167] 646 | lr: 6.250000e-06, epoch: [60400/150000], all_loss: [0.0805], m_loss:[0.0560], s_loss:[0.0245] 647 | lr: 6.250000e-06, epoch: [60500/150000], all_loss: [0.0393], m_loss:[0.0239], s_loss:[0.0154] 648 | lr: 6.250000e-06, epoch: [60600/150000], all_loss: [0.0543], m_loss:[0.0373], s_loss:[0.0170] 649 | lr: 6.250000e-06, epoch: [60700/150000], all_loss: [0.0315], m_loss:[0.0182], s_loss:[0.0134] 650 | lr: 6.250000e-06, epoch: [60800/150000], all_loss: [0.0470], m_loss:[0.0299], s_loss:[0.0171] 651 | lr: 6.250000e-06, epoch: [60900/150000], all_loss: [0.0431], m_loss:[0.0249], s_loss:[0.0182] 652 | lr: 6.250000e-06, epoch: [61000/150000], all_loss: [0.0908], m_loss:[0.0582], s_loss:[0.0326] 653 | lr: 6.250000e-06, epoch: [61100/150000], all_loss: [0.0405], m_loss:[0.0260], s_loss:[0.0146] 654 | lr: 6.250000e-06, epoch: [61200/150000], all_loss: [0.0194], m_loss:[0.0121], s_loss:[0.0073] 655 | lr: 6.250000e-06, epoch: [61300/150000], all_loss: [0.0409], m_loss:[0.0297], s_loss:[0.0111] 656 | lr: 6.250000e-06, epoch: [61400/150000], all_loss: [0.1131], m_loss:[0.0646], s_loss:[0.0485] 657 | lr: 6.250000e-06, epoch: [61500/150000], all_loss: [0.0694], m_loss:[0.0546], s_loss:[0.0148] 658 | lr: 6.250000e-06, epoch: [61600/150000], all_loss: [0.0850], m_loss:[0.0651], s_loss:[0.0200] 659 | lr: 6.250000e-06, epoch: [61700/150000], all_loss: [0.0308], m_loss:[0.0193], s_loss:[0.0115] 660 | lr: 6.250000e-06, epoch: [61800/150000], all_loss: [0.0281], m_loss:[0.0203], s_loss:[0.0077] 661 | lr: 6.250000e-06, epoch: [61900/150000], all_loss: [0.0382], m_loss:[0.0267], s_loss:[0.0115] 662 | lr: 6.250000e-06, epoch: [62000/150000], all_loss: [0.0258], m_loss:[0.0170], s_loss:[0.0088] 663 | lr: 6.250000e-06, epoch: [62100/150000], all_loss: [0.0475], m_loss:[0.0287], s_loss:[0.0188] 664 | lr: 6.250000e-06, epoch: [62200/150000], all_loss: [0.0577], m_loss:[0.0396], s_loss:[0.0181] 665 | lr: 6.250000e-06, epoch: [62300/150000], all_loss: [0.0567], m_loss:[0.0369], s_loss:[0.0198] 666 | lr: 6.250000e-06, epoch: [62400/150000], all_loss: [0.0897], m_loss:[0.0656], s_loss:[0.0241] 667 | lr: 6.250000e-06, epoch: [62500/150000], all_loss: [0.0594], m_loss:[0.0422], s_loss:[0.0172] 668 | iCoseg Cosal2015 f: [0.8435], m: [0.0722], f: [0.8520], m: [0.0805] 669 | lr: 6.250000e-06, epoch: [62600/150000], all_loss: [0.0708], m_loss:[0.0432], s_loss:[0.0276] 670 | lr: 6.250000e-06, epoch: [62700/150000], all_loss: [0.0594], m_loss:[0.0391], s_loss:[0.0203] 671 | lr: 6.250000e-06, epoch: [62800/150000], all_loss: [0.0640], m_loss:[0.0484], s_loss:[0.0156] 672 | lr: 6.250000e-06, epoch: [62900/150000], all_loss: [0.0405], m_loss:[0.0294], s_loss:[0.0112] 673 | lr: 6.250000e-06, epoch: [63000/150000], all_loss: [0.0612], m_loss:[0.0376], s_loss:[0.0236] 674 | lr: 6.250000e-06, epoch: [63100/150000], all_loss: [0.0235], m_loss:[0.0145], s_loss:[0.0090] 675 | lr: 6.250000e-06, epoch: [63200/150000], all_loss: [0.0640], m_loss:[0.0395], s_loss:[0.0245] 676 | lr: 6.250000e-06, epoch: [63300/150000], all_loss: [0.1912], m_loss:[0.1344], s_loss:[0.0568] 677 | lr: 6.250000e-06, epoch: [63400/150000], all_loss: [0.0236], m_loss:[0.0142], s_loss:[0.0094] 678 | lr: 6.250000e-06, epoch: [63500/150000], all_loss: [0.0448], m_loss:[0.0285], s_loss:[0.0163] 679 | lr: 6.250000e-06, epoch: [63600/150000], all_loss: [0.0637], m_loss:[0.0376], s_loss:[0.0262] 680 | lr: 6.250000e-06, epoch: [63700/150000], all_loss: [0.1484], m_loss:[0.1017], s_loss:[0.0467] 681 | lr: 6.250000e-06, epoch: [63800/150000], all_loss: [0.0289], m_loss:[0.0184], s_loss:[0.0105] 682 | lr: 6.250000e-06, epoch: [63900/150000], all_loss: [0.0389], m_loss:[0.0228], s_loss:[0.0161] 683 | lr: 6.250000e-06, epoch: [64000/150000], all_loss: [0.0519], m_loss:[0.0368], s_loss:[0.0150] 684 | lr: 6.250000e-06, epoch: [64100/150000], all_loss: [0.0324], m_loss:[0.0204], s_loss:[0.0120] 685 | lr: 6.250000e-06, epoch: [64200/150000], all_loss: [0.0818], m_loss:[0.0519], s_loss:[0.0299] 686 | lr: 6.250000e-06, epoch: [64300/150000], all_loss: [0.0876], m_loss:[0.0589], s_loss:[0.0287] 687 | lr: 6.250000e-06, epoch: [64400/150000], all_loss: [0.0355], m_loss:[0.0226], s_loss:[0.0129] 688 | lr: 6.250000e-06, epoch: [64500/150000], all_loss: [0.0369], m_loss:[0.0248], s_loss:[0.0121] 689 | lr: 6.250000e-06, epoch: [64600/150000], all_loss: [0.0216], m_loss:[0.0149], s_loss:[0.0067] 690 | lr: 6.250000e-06, epoch: [64700/150000], all_loss: [0.0658], m_loss:[0.0442], s_loss:[0.0216] 691 | lr: 6.250000e-06, epoch: [64800/150000], all_loss: [0.0440], m_loss:[0.0271], s_loss:[0.0168] 692 | lr: 6.250000e-06, epoch: [64900/150000], all_loss: [0.0201], m_loss:[0.0117], s_loss:[0.0084] 693 | lr: 6.250000e-06, epoch: [65000/150000], all_loss: [0.0547], m_loss:[0.0340], s_loss:[0.0207] 694 | iCoseg Cosal2015 f: [0.8418], m: [0.0724], f: [0.8520], m: [0.0791] 695 | lr: 6.250000e-06, epoch: [65100/150000], all_loss: [0.0494], m_loss:[0.0296], s_loss:[0.0197] 696 | lr: 6.250000e-06, epoch: [65200/150000], all_loss: [0.0315], m_loss:[0.0210], s_loss:[0.0104] 697 | lr: 6.250000e-06, epoch: [65300/150000], all_loss: [0.0423], m_loss:[0.0301], s_loss:[0.0122] 698 | lr: 6.250000e-06, epoch: [65400/150000], all_loss: [0.0394], m_loss:[0.0247], s_loss:[0.0147] 699 | lr: 6.250000e-06, epoch: [65500/150000], all_loss: [0.0232], m_loss:[0.0129], s_loss:[0.0103] 700 | lr: 6.250000e-06, epoch: [65600/150000], all_loss: [0.0291], m_loss:[0.0166], s_loss:[0.0125] 701 | lr: 6.250000e-06, epoch: [65700/150000], all_loss: [0.0512], m_loss:[0.0375], s_loss:[0.0137] 702 | lr: 6.250000e-06, epoch: [65800/150000], all_loss: [0.0623], m_loss:[0.0437], s_loss:[0.0186] 703 | lr: 6.250000e-06, epoch: [65900/150000], all_loss: [0.0261], m_loss:[0.0167], s_loss:[0.0095] 704 | lr: 6.250000e-06, epoch: [66000/150000], all_loss: [0.0381], m_loss:[0.0238], s_loss:[0.0143] 705 | lr: 6.250000e-06, epoch: [66100/150000], all_loss: [0.1307], m_loss:[0.0818], s_loss:[0.0489] 706 | lr: 6.250000e-06, epoch: [66200/150000], all_loss: [0.0359], m_loss:[0.0261], s_loss:[0.0098] 707 | lr: 6.250000e-06, epoch: [66300/150000], all_loss: [0.0826], m_loss:[0.0647], s_loss:[0.0179] 708 | lr: 6.250000e-06, epoch: [66400/150000], all_loss: [0.0581], m_loss:[0.0403], s_loss:[0.0178] 709 | lr: 6.250000e-06, epoch: [66500/150000], all_loss: [0.0513], m_loss:[0.0347], s_loss:[0.0166] 710 | lr: 6.250000e-06, epoch: [66600/150000], all_loss: [0.0541], m_loss:[0.0391], s_loss:[0.0150] 711 | lr: 6.250000e-06, epoch: [66700/150000], all_loss: [0.0478], m_loss:[0.0341], s_loss:[0.0137] 712 | lr: 6.250000e-06, epoch: [66800/150000], all_loss: [0.0265], m_loss:[0.0147], s_loss:[0.0118] 713 | lr: 6.250000e-06, epoch: [66900/150000], all_loss: [0.0516], m_loss:[0.0363], s_loss:[0.0153] 714 | lr: 6.250000e-06, epoch: [67000/150000], all_loss: [0.0441], m_loss:[0.0272], s_loss:[0.0170] 715 | lr: 6.250000e-06, epoch: [67100/150000], all_loss: [0.0328], m_loss:[0.0186], s_loss:[0.0142] 716 | lr: 6.250000e-06, epoch: [67200/150000], all_loss: [0.0466], m_loss:[0.0291], s_loss:[0.0175] 717 | lr: 6.250000e-06, epoch: [67300/150000], all_loss: [0.0410], m_loss:[0.0253], s_loss:[0.0157] 718 | lr: 6.250000e-06, epoch: [67400/150000], all_loss: [0.0295], m_loss:[0.0175], s_loss:[0.0120] 719 | lr: 6.250000e-06, epoch: [67500/150000], all_loss: [0.0157], m_loss:[0.0088], s_loss:[0.0069] 720 | iCoseg Cosal2015 f: [0.8428], m: [0.0728], f: [0.8530], m: [0.0792] 721 | lr: 6.250000e-06, epoch: [67600/150000], all_loss: [0.0489], m_loss:[0.0304], s_loss:[0.0186] 722 | lr: 6.250000e-06, epoch: [67700/150000], all_loss: [0.0426], m_loss:[0.0287], s_loss:[0.0139] 723 | lr: 6.250000e-06, epoch: [67800/150000], all_loss: [0.0408], m_loss:[0.0277], s_loss:[0.0131] 724 | lr: 6.250000e-06, epoch: [67900/150000], all_loss: [0.1162], m_loss:[0.0827], s_loss:[0.0334] 725 | lr: 6.250000e-06, epoch: [68000/150000], all_loss: [0.0250], m_loss:[0.0159], s_loss:[0.0091] 726 | lr: 6.250000e-06, epoch: [68100/150000], all_loss: [0.0297], m_loss:[0.0172], s_loss:[0.0125] 727 | lr: 6.250000e-06, epoch: [68200/150000], all_loss: [0.0351], m_loss:[0.0226], s_loss:[0.0126] 728 | lr: 6.250000e-06, epoch: [68300/150000], all_loss: [0.0467], m_loss:[0.0306], s_loss:[0.0161] 729 | lr: 6.250000e-06, epoch: [68400/150000], all_loss: [0.0447], m_loss:[0.0317], s_loss:[0.0130] 730 | lr: 6.250000e-06, epoch: [68500/150000], all_loss: [0.0369], m_loss:[0.0245], s_loss:[0.0124] 731 | lr: 6.250000e-06, epoch: [68600/150000], all_loss: [0.0270], m_loss:[0.0175], s_loss:[0.0094] 732 | lr: 6.250000e-06, epoch: [68700/150000], all_loss: [0.0572], m_loss:[0.0437], s_loss:[0.0135] 733 | lr: 6.250000e-06, epoch: [68800/150000], all_loss: [0.0476], m_loss:[0.0347], s_loss:[0.0129] 734 | lr: 6.250000e-06, epoch: [68900/150000], all_loss: [0.0349], m_loss:[0.0230], s_loss:[0.0118] 735 | lr: 6.250000e-06, epoch: [69000/150000], all_loss: [0.0715], m_loss:[0.0397], s_loss:[0.0318] 736 | lr: 6.250000e-06, epoch: [69100/150000], all_loss: [0.0411], m_loss:[0.0294], s_loss:[0.0116] 737 | lr: 6.250000e-06, epoch: [69200/150000], all_loss: [0.0203], m_loss:[0.0116], s_loss:[0.0087] 738 | lr: 6.250000e-06, epoch: [69300/150000], all_loss: [0.0425], m_loss:[0.0293], s_loss:[0.0132] 739 | lr: 6.250000e-06, epoch: [69400/150000], all_loss: [0.0358], m_loss:[0.0228], s_loss:[0.0130] 740 | lr: 6.250000e-06, epoch: [69500/150000], all_loss: [0.0541], m_loss:[0.0327], s_loss:[0.0214] 741 | lr: 6.250000e-06, epoch: [69600/150000], all_loss: [0.0574], m_loss:[0.0381], s_loss:[0.0192] 742 | lr: 6.250000e-06, epoch: [69700/150000], all_loss: [0.0447], m_loss:[0.0338], s_loss:[0.0109] 743 | lr: 6.250000e-06, epoch: [69800/150000], all_loss: [0.0333], m_loss:[0.0234], s_loss:[0.0099] 744 | lr: 6.250000e-06, epoch: [69900/150000], all_loss: [0.0489], m_loss:[0.0337], s_loss:[0.0152] 745 | lr: 6.250000e-06, epoch: [70000/150000], all_loss: [0.0573], m_loss:[0.0385], s_loss:[0.0188] 746 | iCoseg Cosal2015 f: [0.8418], m: [0.0757], f: [0.8532], m: [0.0824] 747 | lr: 6.250000e-06, epoch: [70100/150000], all_loss: [0.0735], m_loss:[0.0513], s_loss:[0.0222] 748 | lr: 6.250000e-06, epoch: [70200/150000], all_loss: [0.0602], m_loss:[0.0368], s_loss:[0.0234] 749 | lr: 6.250000e-06, epoch: [70300/150000], all_loss: [0.0647], m_loss:[0.0435], s_loss:[0.0212] 750 | lr: 6.250000e-06, epoch: [70400/150000], all_loss: [0.0769], m_loss:[0.0571], s_loss:[0.0198] 751 | lr: 6.250000e-06, epoch: [70500/150000], all_loss: [0.0343], m_loss:[0.0176], s_loss:[0.0167] 752 | lr: 6.250000e-06, epoch: [70600/150000], all_loss: [0.0308], m_loss:[0.0200], s_loss:[0.0108] 753 | lr: 6.250000e-06, epoch: [70700/150000], all_loss: [0.0407], m_loss:[0.0255], s_loss:[0.0153] 754 | lr: 6.250000e-06, epoch: [70800/150000], all_loss: [0.0353], m_loss:[0.0219], s_loss:[0.0134] 755 | lr: 6.250000e-06, epoch: [70900/150000], all_loss: [0.0287], m_loss:[0.0154], s_loss:[0.0133] 756 | lr: 6.250000e-06, epoch: [71000/150000], all_loss: [0.0395], m_loss:[0.0280], s_loss:[0.0116] 757 | lr: 6.250000e-06, epoch: [71100/150000], all_loss: [0.0443], m_loss:[0.0293], s_loss:[0.0150] 758 | lr: 6.250000e-06, epoch: [71200/150000], all_loss: [0.1189], m_loss:[0.0834], s_loss:[0.0355] 759 | lr: 6.250000e-06, epoch: [71300/150000], all_loss: [0.0506], m_loss:[0.0327], s_loss:[0.0179] 760 | lr: 6.250000e-06, epoch: [71400/150000], all_loss: [0.0558], m_loss:[0.0349], s_loss:[0.0209] 761 | lr: 6.250000e-06, epoch: [71500/150000], all_loss: [0.0454], m_loss:[0.0273], s_loss:[0.0180] 762 | lr: 6.250000e-06, epoch: [71600/150000], all_loss: [0.1063], m_loss:[0.0814], s_loss:[0.0249] 763 | lr: 6.250000e-06, epoch: [71700/150000], all_loss: [0.0556], m_loss:[0.0355], s_loss:[0.0202] 764 | lr: 6.250000e-06, epoch: [71800/150000], all_loss: [0.0372], m_loss:[0.0212], s_loss:[0.0160] 765 | lr: 6.250000e-06, epoch: [71900/150000], all_loss: [0.0189], m_loss:[0.0101], s_loss:[0.0088] 766 | lr: 6.250000e-06, epoch: [72000/150000], all_loss: [0.0749], m_loss:[0.0496], s_loss:[0.0253] 767 | lr: 6.250000e-06, epoch: [72100/150000], all_loss: [0.0320], m_loss:[0.0233], s_loss:[0.0087] 768 | lr: 6.250000e-06, epoch: [72200/150000], all_loss: [0.0294], m_loss:[0.0179], s_loss:[0.0114] 769 | lr: 6.250000e-06, epoch: [72300/150000], all_loss: [0.0550], m_loss:[0.0322], s_loss:[0.0228] 770 | lr: 6.250000e-06, epoch: [72400/150000], all_loss: [0.1052], m_loss:[0.0625], s_loss:[0.0426] 771 | lr: 6.250000e-06, epoch: [72500/150000], all_loss: [0.0560], m_loss:[0.0402], s_loss:[0.0157] 772 | iCoseg Cosal2015 f: [0.8431], m: [0.0720], f: [0.8517], m: [0.0791] 773 | lr: 6.250000e-06, epoch: [72600/150000], all_loss: [0.0411], m_loss:[0.0280], s_loss:[0.0131] 774 | lr: 6.250000e-06, epoch: [72700/150000], all_loss: [0.0765], m_loss:[0.0501], s_loss:[0.0264] 775 | lr: 6.250000e-06, epoch: [72800/150000], all_loss: [0.0428], m_loss:[0.0322], s_loss:[0.0106] 776 | lr: 6.250000e-06, epoch: [72900/150000], all_loss: [0.0428], m_loss:[0.0259], s_loss:[0.0169] 777 | lr: 6.250000e-06, epoch: [73000/150000], all_loss: [0.0804], m_loss:[0.0576], s_loss:[0.0228] 778 | lr: 6.250000e-06, epoch: [73100/150000], all_loss: [0.1161], m_loss:[0.0746], s_loss:[0.0415] 779 | lr: 6.250000e-06, epoch: [73200/150000], all_loss: [0.0342], m_loss:[0.0209], s_loss:[0.0132] 780 | lr: 6.250000e-06, epoch: [73300/150000], all_loss: [0.0483], m_loss:[0.0355], s_loss:[0.0128] 781 | lr: 6.250000e-06, epoch: [73400/150000], all_loss: [0.0237], m_loss:[0.0165], s_loss:[0.0072] 782 | lr: 6.250000e-06, epoch: [73500/150000], all_loss: [0.0502], m_loss:[0.0356], s_loss:[0.0146] 783 | lr: 6.250000e-06, epoch: [73600/150000], all_loss: [0.1307], m_loss:[0.0867], s_loss:[0.0440] 784 | lr: 6.250000e-06, epoch: [73700/150000], all_loss: [0.0515], m_loss:[0.0302], s_loss:[0.0212] 785 | lr: 6.250000e-06, epoch: [73800/150000], all_loss: [0.0447], m_loss:[0.0297], s_loss:[0.0150] 786 | lr: 6.250000e-06, epoch: [73900/150000], all_loss: [0.0243], m_loss:[0.0147], s_loss:[0.0095] 787 | lr: 6.250000e-06, epoch: [74000/150000], all_loss: [0.0350], m_loss:[0.0223], s_loss:[0.0128] 788 | lr: 6.250000e-06, epoch: [74100/150000], all_loss: [0.0286], m_loss:[0.0155], s_loss:[0.0131] 789 | lr: 6.250000e-06, epoch: [74200/150000], all_loss: [0.0802], m_loss:[0.0520], s_loss:[0.0282] 790 | lr: 6.250000e-06, epoch: [74300/150000], all_loss: [0.0426], m_loss:[0.0287], s_loss:[0.0139] 791 | lr: 6.250000e-06, epoch: [74400/150000], all_loss: [0.0472], m_loss:[0.0334], s_loss:[0.0138] 792 | lr: 6.250000e-06, epoch: [74500/150000], all_loss: [0.0417], m_loss:[0.0280], s_loss:[0.0137] 793 | lr: 6.250000e-06, epoch: [74600/150000], all_loss: [0.0258], m_loss:[0.0159], s_loss:[0.0100] 794 | lr: 6.250000e-06, epoch: [74700/150000], all_loss: [0.0795], m_loss:[0.0480], s_loss:[0.0315] 795 | lr: 6.250000e-06, epoch: [74800/150000], all_loss: [0.1226], m_loss:[0.0738], s_loss:[0.0487] 796 | lr: 6.250000e-06, epoch: [74900/150000], all_loss: [0.1008], m_loss:[0.0715], s_loss:[0.0293] 797 | lr: 6.250000e-06, epoch: [75000/150000], all_loss: [0.0683], m_loss:[0.0448], s_loss:[0.0235] 798 | iCoseg Cosal2015 f: [0.8413], m: [0.0732], f: [0.8521], m: [0.0769] 799 | lr: 3.125000e-06, epoch: [75100/150000], all_loss: [0.1648], m_loss:[0.1277], s_loss:[0.0371] 800 | lr: 3.125000e-06, epoch: [75200/150000], all_loss: [0.0258], m_loss:[0.0182], s_loss:[0.0076] 801 | lr: 3.125000e-06, epoch: [75300/150000], all_loss: [0.0406], m_loss:[0.0227], s_loss:[0.0179] 802 | lr: 3.125000e-06, epoch: [75400/150000], all_loss: [0.0474], m_loss:[0.0272], s_loss:[0.0202] 803 | lr: 3.125000e-06, epoch: [75500/150000], all_loss: [0.0400], m_loss:[0.0205], s_loss:[0.0195] 804 | lr: 3.125000e-06, epoch: [75600/150000], all_loss: [0.2625], m_loss:[0.1862], s_loss:[0.0763] 805 | lr: 3.125000e-06, epoch: [75700/150000], all_loss: [0.0414], m_loss:[0.0277], s_loss:[0.0137] 806 | lr: 3.125000e-06, epoch: [75800/150000], all_loss: [0.0852], m_loss:[0.0498], s_loss:[0.0354] 807 | lr: 3.125000e-06, epoch: [75900/150000], all_loss: [0.0404], m_loss:[0.0240], s_loss:[0.0164] 808 | lr: 3.125000e-06, epoch: [76000/150000], all_loss: [0.0258], m_loss:[0.0149], s_loss:[0.0109] 809 | lr: 3.125000e-06, epoch: [76100/150000], all_loss: [0.0120], m_loss:[0.0070], s_loss:[0.0050] 810 | lr: 3.125000e-06, epoch: [76200/150000], all_loss: [0.0268], m_loss:[0.0149], s_loss:[0.0119] 811 | lr: 3.125000e-06, epoch: [76300/150000], all_loss: [0.0893], m_loss:[0.0583], s_loss:[0.0310] 812 | lr: 3.125000e-06, epoch: [76400/150000], all_loss: [0.0428], m_loss:[0.0292], s_loss:[0.0136] 813 | lr: 3.125000e-06, epoch: [76500/150000], all_loss: [0.0272], m_loss:[0.0193], s_loss:[0.0080] 814 | lr: 3.125000e-06, epoch: [76600/150000], all_loss: [0.0535], m_loss:[0.0314], s_loss:[0.0221] 815 | lr: 3.125000e-06, epoch: [76700/150000], all_loss: [0.0290], m_loss:[0.0191], s_loss:[0.0099] 816 | lr: 3.125000e-06, epoch: [76800/150000], all_loss: [0.0694], m_loss:[0.0458], s_loss:[0.0236] 817 | lr: 3.125000e-06, epoch: [76900/150000], all_loss: [0.0158], m_loss:[0.0080], s_loss:[0.0078] 818 | lr: 3.125000e-06, epoch: [77000/150000], all_loss: [0.0917], m_loss:[0.0602], s_loss:[0.0316] 819 | lr: 3.125000e-06, epoch: [77100/150000], all_loss: [0.0394], m_loss:[0.0264], s_loss:[0.0130] 820 | lr: 3.125000e-06, epoch: [77200/150000], all_loss: [0.0538], m_loss:[0.0380], s_loss:[0.0159] 821 | lr: 3.125000e-06, epoch: [77300/150000], all_loss: [0.0454], m_loss:[0.0346], s_loss:[0.0108] 822 | lr: 3.125000e-06, epoch: [77400/150000], all_loss: [0.0434], m_loss:[0.0275], s_loss:[0.0159] 823 | lr: 3.125000e-06, epoch: [77500/150000], all_loss: [0.0273], m_loss:[0.0159], s_loss:[0.0114] 824 | iCoseg Cosal2015 f: [0.8428], m: [0.0729], f: [0.8534], m: [0.0783] 825 | lr: 3.125000e-06, epoch: [77600/150000], all_loss: [0.1125], m_loss:[0.0694], s_loss:[0.0431] 826 | lr: 3.125000e-06, epoch: [77700/150000], all_loss: [0.0380], m_loss:[0.0251], s_loss:[0.0130] 827 | lr: 3.125000e-06, epoch: [77800/150000], all_loss: [0.0569], m_loss:[0.0392], s_loss:[0.0177] 828 | lr: 3.125000e-06, epoch: [77900/150000], all_loss: [0.0505], m_loss:[0.0279], s_loss:[0.0227] 829 | lr: 3.125000e-06, epoch: [78000/150000], all_loss: [0.0283], m_loss:[0.0177], s_loss:[0.0107] 830 | lr: 3.125000e-06, epoch: [78100/150000], all_loss: [0.0516], m_loss:[0.0334], s_loss:[0.0182] 831 | lr: 3.125000e-06, epoch: [78200/150000], all_loss: [0.0375], m_loss:[0.0231], s_loss:[0.0144] 832 | lr: 3.125000e-06, epoch: [78300/150000], all_loss: [0.0269], m_loss:[0.0134], s_loss:[0.0135] 833 | lr: 3.125000e-06, epoch: [78400/150000], all_loss: [0.1174], m_loss:[0.0879], s_loss:[0.0295] 834 | lr: 3.125000e-06, epoch: [78500/150000], all_loss: [0.1935], m_loss:[0.1369], s_loss:[0.0567] 835 | lr: 3.125000e-06, epoch: [78600/150000], all_loss: [0.0999], m_loss:[0.0606], s_loss:[0.0393] 836 | lr: 3.125000e-06, epoch: [78700/150000], all_loss: [0.0721], m_loss:[0.0567], s_loss:[0.0154] 837 | lr: 3.125000e-06, epoch: [78800/150000], all_loss: [0.0790], m_loss:[0.0587], s_loss:[0.0204] 838 | lr: 3.125000e-06, epoch: [78900/150000], all_loss: [0.0249], m_loss:[0.0160], s_loss:[0.0089] 839 | lr: 3.125000e-06, epoch: [79000/150000], all_loss: [0.0417], m_loss:[0.0285], s_loss:[0.0132] 840 | lr: 3.125000e-06, epoch: [79100/150000], all_loss: [0.0903], m_loss:[0.0640], s_loss:[0.0264] 841 | lr: 3.125000e-06, epoch: [79200/150000], all_loss: [0.0304], m_loss:[0.0210], s_loss:[0.0094] 842 | lr: 3.125000e-06, epoch: [79300/150000], all_loss: [0.0457], m_loss:[0.0299], s_loss:[0.0157] 843 | lr: 3.125000e-06, epoch: [79400/150000], all_loss: [0.0352], m_loss:[0.0251], s_loss:[0.0101] 844 | lr: 3.125000e-06, epoch: [79500/150000], all_loss: [0.0240], m_loss:[0.0151], s_loss:[0.0090] 845 | lr: 3.125000e-06, epoch: [79600/150000], all_loss: [0.0390], m_loss:[0.0235], s_loss:[0.0156] 846 | lr: 3.125000e-06, epoch: [79700/150000], all_loss: [0.0343], m_loss:[0.0194], s_loss:[0.0149] 847 | lr: 3.125000e-06, epoch: [79800/150000], all_loss: [0.0406], m_loss:[0.0244], s_loss:[0.0163] 848 | lr: 3.125000e-06, epoch: [79900/150000], all_loss: [0.0201], m_loss:[0.0128], s_loss:[0.0074] 849 | lr: 3.125000e-06, epoch: [80000/150000], all_loss: [0.0494], m_loss:[0.0332], s_loss:[0.0162] 850 | iCoseg Cosal2015 f: [0.8417], m: [0.0740], f: [0.8525], m: [0.0809] 851 | lr: 3.125000e-06, epoch: [80100/150000], all_loss: [0.0914], m_loss:[0.0677], s_loss:[0.0237] 852 | lr: 3.125000e-06, epoch: [80200/150000], all_loss: [0.0475], m_loss:[0.0297], s_loss:[0.0178] 853 | lr: 3.125000e-06, epoch: [80300/150000], all_loss: [0.1430], m_loss:[0.0918], s_loss:[0.0512] 854 | lr: 3.125000e-06, epoch: [80400/150000], all_loss: [0.0414], m_loss:[0.0272], s_loss:[0.0143] 855 | lr: 3.125000e-06, epoch: [80500/150000], all_loss: [0.0716], m_loss:[0.0477], s_loss:[0.0239] 856 | lr: 3.125000e-06, epoch: [80600/150000], all_loss: [0.0554], m_loss:[0.0363], s_loss:[0.0191] 857 | lr: 3.125000e-06, epoch: [80700/150000], all_loss: [0.0460], m_loss:[0.0288], s_loss:[0.0171] 858 | lr: 3.125000e-06, epoch: [80800/150000], all_loss: [0.0450], m_loss:[0.0342], s_loss:[0.0108] 859 | lr: 3.125000e-06, epoch: [80900/150000], all_loss: [0.1855], m_loss:[0.1258], s_loss:[0.0597] 860 | lr: 3.125000e-06, epoch: [81000/150000], all_loss: [0.0454], m_loss:[0.0311], s_loss:[0.0143] 861 | lr: 3.125000e-06, epoch: [81100/150000], all_loss: [0.0372], m_loss:[0.0272], s_loss:[0.0100] 862 | lr: 3.125000e-06, epoch: [81200/150000], all_loss: [0.0221], m_loss:[0.0147], s_loss:[0.0074] 863 | lr: 3.125000e-06, epoch: [81300/150000], all_loss: [0.0492], m_loss:[0.0332], s_loss:[0.0160] 864 | lr: 3.125000e-06, epoch: [81400/150000], all_loss: [0.0516], m_loss:[0.0277], s_loss:[0.0239] 865 | lr: 3.125000e-06, epoch: [81500/150000], all_loss: [0.0169], m_loss:[0.0098], s_loss:[0.0071] 866 | lr: 3.125000e-06, epoch: [81600/150000], all_loss: [0.0332], m_loss:[0.0213], s_loss:[0.0118] 867 | lr: 3.125000e-06, epoch: [81700/150000], all_loss: [0.0290], m_loss:[0.0177], s_loss:[0.0113] 868 | lr: 3.125000e-06, epoch: [81800/150000], all_loss: [0.0595], m_loss:[0.0345], s_loss:[0.0249] 869 | lr: 3.125000e-06, epoch: [81900/150000], all_loss: [0.0252], m_loss:[0.0157], s_loss:[0.0094] 870 | lr: 3.125000e-06, epoch: [82000/150000], all_loss: [0.0378], m_loss:[0.0277], s_loss:[0.0102] 871 | lr: 3.125000e-06, epoch: [82100/150000], all_loss: [0.0341], m_loss:[0.0233], s_loss:[0.0108] 872 | lr: 3.125000e-06, epoch: [82200/150000], all_loss: [0.0675], m_loss:[0.0426], s_loss:[0.0249] 873 | lr: 3.125000e-06, epoch: [82300/150000], all_loss: [0.0682], m_loss:[0.0486], s_loss:[0.0196] 874 | lr: 3.125000e-06, epoch: [82400/150000], all_loss: [0.2640], m_loss:[0.1748], s_loss:[0.0892] 875 | lr: 3.125000e-06, epoch: [82500/150000], all_loss: [0.0837], m_loss:[0.0557], s_loss:[0.0280] 876 | iCoseg Cosal2015 f: [0.8405], m: [0.0725], f: [0.8518], m: [0.0791] 877 | lr: 3.125000e-06, epoch: [82600/150000], all_loss: [0.0551], m_loss:[0.0334], s_loss:[0.0217] 878 | lr: 3.125000e-06, epoch: [82700/150000], all_loss: [0.0263], m_loss:[0.0192], s_loss:[0.0072] 879 | lr: 3.125000e-06, epoch: [82800/150000], all_loss: [0.0417], m_loss:[0.0276], s_loss:[0.0141] 880 | lr: 3.125000e-06, epoch: [82900/150000], all_loss: [0.0254], m_loss:[0.0156], s_loss:[0.0098] 881 | lr: 3.125000e-06, epoch: [83000/150000], all_loss: [0.0568], m_loss:[0.0357], s_loss:[0.0210] 882 | lr: 3.125000e-06, epoch: [83100/150000], all_loss: [0.0416], m_loss:[0.0280], s_loss:[0.0136] 883 | lr: 3.125000e-06, epoch: [83200/150000], all_loss: [0.0590], m_loss:[0.0421], s_loss:[0.0169] 884 | lr: 3.125000e-06, epoch: [83300/150000], all_loss: [0.0336], m_loss:[0.0199], s_loss:[0.0137] 885 | lr: 3.125000e-06, epoch: [83400/150000], all_loss: [0.0432], m_loss:[0.0260], s_loss:[0.0173] 886 | lr: 3.125000e-06, epoch: [83500/150000], all_loss: [0.0252], m_loss:[0.0154], s_loss:[0.0097] 887 | lr: 3.125000e-06, epoch: [83600/150000], all_loss: [0.0412], m_loss:[0.0298], s_loss:[0.0114] 888 | lr: 3.125000e-06, epoch: [83700/150000], all_loss: [0.0477], m_loss:[0.0335], s_loss:[0.0142] 889 | lr: 3.125000e-06, epoch: [83800/150000], all_loss: [0.0738], m_loss:[0.0490], s_loss:[0.0248] 890 | lr: 3.125000e-06, epoch: [83900/150000], all_loss: [0.0240], m_loss:[0.0141], s_loss:[0.0099] 891 | lr: 3.125000e-06, epoch: [84000/150000], all_loss: [0.0382], m_loss:[0.0260], s_loss:[0.0122] 892 | lr: 3.125000e-06, epoch: [84100/150000], all_loss: [0.0629], m_loss:[0.0385], s_loss:[0.0244] 893 | lr: 3.125000e-06, epoch: [84200/150000], all_loss: [0.0330], m_loss:[0.0198], s_loss:[0.0132] 894 | lr: 3.125000e-06, epoch: [84300/150000], all_loss: [0.0291], m_loss:[0.0202], s_loss:[0.0089] 895 | lr: 3.125000e-06, epoch: [84400/150000], all_loss: [0.1136], m_loss:[0.0788], s_loss:[0.0349] 896 | lr: 3.125000e-06, epoch: [84500/150000], all_loss: [0.0891], m_loss:[0.0595], s_loss:[0.0296] 897 | lr: 3.125000e-06, epoch: [84600/150000], all_loss: [0.0531], m_loss:[0.0362], s_loss:[0.0169] 898 | lr: 3.125000e-06, epoch: [84700/150000], all_loss: [0.0426], m_loss:[0.0283], s_loss:[0.0143] 899 | lr: 3.125000e-06, epoch: [84800/150000], all_loss: [0.0961], m_loss:[0.0660], s_loss:[0.0301] 900 | lr: 3.125000e-06, epoch: [84900/150000], all_loss: [0.0243], m_loss:[0.0161], s_loss:[0.0083] 901 | lr: 3.125000e-06, epoch: [85000/150000], all_loss: [0.0270], m_loss:[0.0179], s_loss:[0.0092] 902 | iCoseg Cosal2015 f: [0.8404], m: [0.0733], f: [0.8532], m: [0.0788] 903 | lr: 3.125000e-06, epoch: [85100/150000], all_loss: [0.0442], m_loss:[0.0334], s_loss:[0.0108] 904 | lr: 3.125000e-06, epoch: [85200/150000], all_loss: [0.0365], m_loss:[0.0256], s_loss:[0.0109] 905 | lr: 3.125000e-06, epoch: [85300/150000], all_loss: [0.0392], m_loss:[0.0260], s_loss:[0.0133] 906 | lr: 3.125000e-06, epoch: [85400/150000], all_loss: [0.0563], m_loss:[0.0335], s_loss:[0.0228] 907 | lr: 3.125000e-06, epoch: [85500/150000], all_loss: [0.1312], m_loss:[0.0916], s_loss:[0.0396] 908 | lr: 3.125000e-06, epoch: [85600/150000], all_loss: [0.0364], m_loss:[0.0213], s_loss:[0.0151] 909 | lr: 3.125000e-06, epoch: [85700/150000], all_loss: [0.0461], m_loss:[0.0325], s_loss:[0.0136] 910 | lr: 3.125000e-06, epoch: [85800/150000], all_loss: [0.1043], m_loss:[0.0754], s_loss:[0.0289] 911 | lr: 3.125000e-06, epoch: [85900/150000], all_loss: [0.0236], m_loss:[0.0170], s_loss:[0.0066] 912 | lr: 3.125000e-06, epoch: [86000/150000], all_loss: [0.0398], m_loss:[0.0289], s_loss:[0.0109] 913 | lr: 3.125000e-06, epoch: [86100/150000], all_loss: [0.0373], m_loss:[0.0241], s_loss:[0.0132] 914 | lr: 3.125000e-06, epoch: [86200/150000], all_loss: [0.0219], m_loss:[0.0142], s_loss:[0.0077] 915 | lr: 3.125000e-06, epoch: [86300/150000], all_loss: [0.0607], m_loss:[0.0439], s_loss:[0.0167] 916 | lr: 3.125000e-06, epoch: [86400/150000], all_loss: [0.1133], m_loss:[0.0733], s_loss:[0.0400] 917 | lr: 3.125000e-06, epoch: [86500/150000], all_loss: [0.0422], m_loss:[0.0234], s_loss:[0.0188] 918 | lr: 3.125000e-06, epoch: [86600/150000], all_loss: [0.0258], m_loss:[0.0178], s_loss:[0.0080] 919 | lr: 3.125000e-06, epoch: [86700/150000], all_loss: [0.0310], m_loss:[0.0197], s_loss:[0.0113] 920 | lr: 3.125000e-06, epoch: [86800/150000], all_loss: [0.0565], m_loss:[0.0368], s_loss:[0.0197] 921 | lr: 3.125000e-06, epoch: [86900/150000], all_loss: [0.0484], m_loss:[0.0286], s_loss:[0.0198] 922 | lr: 3.125000e-06, epoch: [87000/150000], all_loss: [0.0315], m_loss:[0.0180], s_loss:[0.0135] 923 | lr: 3.125000e-06, epoch: [87100/150000], all_loss: [0.0700], m_loss:[0.0491], s_loss:[0.0209] 924 | lr: 3.125000e-06, epoch: [87200/150000], all_loss: [0.0396], m_loss:[0.0289], s_loss:[0.0107] 925 | lr: 3.125000e-06, epoch: [87300/150000], all_loss: [0.0261], m_loss:[0.0175], s_loss:[0.0086] 926 | lr: 3.125000e-06, epoch: [87400/150000], all_loss: [0.0406], m_loss:[0.0248], s_loss:[0.0158] 927 | lr: 3.125000e-06, epoch: [87500/150000], all_loss: [0.0302], m_loss:[0.0188], s_loss:[0.0114] 928 | iCoseg Cosal2015 f: [0.8406], m: [0.0741], f: [0.8524], m: [0.0817] 929 | lr: 3.125000e-06, epoch: [87600/150000], all_loss: [0.0352], m_loss:[0.0203], s_loss:[0.0149] 930 | lr: 3.125000e-06, epoch: [87700/150000], all_loss: [0.0314], m_loss:[0.0168], s_loss:[0.0145] 931 | lr: 3.125000e-06, epoch: [87800/150000], all_loss: [0.0358], m_loss:[0.0213], s_loss:[0.0145] 932 | lr: 3.125000e-06, epoch: [87900/150000], all_loss: [0.0451], m_loss:[0.0312], s_loss:[0.0138] 933 | lr: 3.125000e-06, epoch: [88000/150000], all_loss: [0.0590], m_loss:[0.0442], s_loss:[0.0148] 934 | lr: 3.125000e-06, epoch: [88100/150000], all_loss: [0.0379], m_loss:[0.0246], s_loss:[0.0133] 935 | lr: 3.125000e-06, epoch: [88200/150000], all_loss: [0.0506], m_loss:[0.0308], s_loss:[0.0198] 936 | lr: 3.125000e-06, epoch: [88300/150000], all_loss: [0.0441], m_loss:[0.0280], s_loss:[0.0161] 937 | lr: 3.125000e-06, epoch: [88400/150000], all_loss: [0.0265], m_loss:[0.0170], s_loss:[0.0095] 938 | lr: 3.125000e-06, epoch: [88500/150000], all_loss: [0.0397], m_loss:[0.0232], s_loss:[0.0165] 939 | lr: 3.125000e-06, epoch: [88600/150000], all_loss: [0.0309], m_loss:[0.0212], s_loss:[0.0098] 940 | lr: 3.125000e-06, epoch: [88700/150000], all_loss: [0.0631], m_loss:[0.0461], s_loss:[0.0170] 941 | lr: 3.125000e-06, epoch: [88800/150000], all_loss: [0.0287], m_loss:[0.0187], s_loss:[0.0100] 942 | lr: 3.125000e-06, epoch: [88900/150000], all_loss: [0.0610], m_loss:[0.0396], s_loss:[0.0214] 943 | lr: 3.125000e-06, epoch: [89000/150000], all_loss: [0.0687], m_loss:[0.0499], s_loss:[0.0188] 944 | lr: 3.125000e-06, epoch: [89100/150000], all_loss: [0.0139], m_loss:[0.0084], s_loss:[0.0055] 945 | lr: 3.125000e-06, epoch: [89200/150000], all_loss: [0.0962], m_loss:[0.0626], s_loss:[0.0336] 946 | lr: 3.125000e-06, epoch: [89300/150000], all_loss: [0.0381], m_loss:[0.0232], s_loss:[0.0150] 947 | lr: 3.125000e-06, epoch: [89400/150000], all_loss: [0.1611], m_loss:[0.1039], s_loss:[0.0572] 948 | lr: 3.125000e-06, epoch: [89500/150000], all_loss: [0.0350], m_loss:[0.0235], s_loss:[0.0115] 949 | lr: 3.125000e-06, epoch: [89600/150000], all_loss: [0.0491], m_loss:[0.0322], s_loss:[0.0169] 950 | lr: 3.125000e-06, epoch: [89700/150000], all_loss: [0.1028], m_loss:[0.0696], s_loss:[0.0332] 951 | lr: 3.125000e-06, epoch: [89800/150000], all_loss: [0.0729], m_loss:[0.0475], s_loss:[0.0253] 952 | lr: 3.125000e-06, epoch: [89900/150000], all_loss: [0.0330], m_loss:[0.0204], s_loss:[0.0126] 953 | lr: 3.125000e-06, epoch: [90000/150000], all_loss: [0.1001], m_loss:[0.0668], s_loss:[0.0333] 954 | iCoseg Cosal2015 f: [0.8415], m: [0.0734], f: [0.8536], m: [0.0796] 955 | lr: 3.125000e-06, epoch: [90100/150000], all_loss: [0.0328], m_loss:[0.0232], s_loss:[0.0097] 956 | lr: 3.125000e-06, epoch: [90200/150000], all_loss: [0.0868], m_loss:[0.0592], s_loss:[0.0276] 957 | lr: 3.125000e-06, epoch: [90300/150000], all_loss: [0.0426], m_loss:[0.0279], s_loss:[0.0147] 958 | lr: 3.125000e-06, epoch: [90400/150000], all_loss: [0.0308], m_loss:[0.0184], s_loss:[0.0123] 959 | lr: 3.125000e-06, epoch: [90500/150000], all_loss: [0.0355], m_loss:[0.0208], s_loss:[0.0147] 960 | lr: 3.125000e-06, epoch: [90600/150000], all_loss: [0.0386], m_loss:[0.0238], s_loss:[0.0148] 961 | lr: 3.125000e-06, epoch: [90700/150000], all_loss: [0.0349], m_loss:[0.0242], s_loss:[0.0106] 962 | lr: 3.125000e-06, epoch: [90800/150000], all_loss: [0.0515], m_loss:[0.0352], s_loss:[0.0163] 963 | lr: 3.125000e-06, epoch: [90900/150000], all_loss: [0.0605], m_loss:[0.0451], s_loss:[0.0154] 964 | lr: 3.125000e-06, epoch: [91000/150000], all_loss: [0.0356], m_loss:[0.0245], s_loss:[0.0111] 965 | lr: 3.125000e-06, epoch: [91100/150000], all_loss: [0.0338], m_loss:[0.0195], s_loss:[0.0144] 966 | lr: 3.125000e-06, epoch: [91200/150000], all_loss: [0.0414], m_loss:[0.0282], s_loss:[0.0131] 967 | lr: 3.125000e-06, epoch: [91300/150000], all_loss: [0.0645], m_loss:[0.0370], s_loss:[0.0275] 968 | lr: 3.125000e-06, epoch: [91400/150000], all_loss: [0.0738], m_loss:[0.0444], s_loss:[0.0294] 969 | lr: 3.125000e-06, epoch: [91500/150000], all_loss: [0.0361], m_loss:[0.0240], s_loss:[0.0121] 970 | lr: 3.125000e-06, epoch: [91600/150000], all_loss: [0.0407], m_loss:[0.0281], s_loss:[0.0126] 971 | lr: 3.125000e-06, epoch: [91700/150000], all_loss: [0.0369], m_loss:[0.0242], s_loss:[0.0127] 972 | lr: 3.125000e-06, epoch: [91800/150000], all_loss: [0.0404], m_loss:[0.0263], s_loss:[0.0141] 973 | lr: 3.125000e-06, epoch: [91900/150000], all_loss: [0.0779], m_loss:[0.0547], s_loss:[0.0232] 974 | lr: 3.125000e-06, epoch: [92000/150000], all_loss: [0.0499], m_loss:[0.0298], s_loss:[0.0201] 975 | lr: 3.125000e-06, epoch: [92100/150000], all_loss: [0.0739], m_loss:[0.0385], s_loss:[0.0354] 976 | lr: 3.125000e-06, epoch: [92200/150000], all_loss: [0.0349], m_loss:[0.0229], s_loss:[0.0120] 977 | lr: 3.125000e-06, epoch: [92300/150000], all_loss: [0.0362], m_loss:[0.0233], s_loss:[0.0129] 978 | lr: 3.125000e-06, epoch: [92400/150000], all_loss: [0.0294], m_loss:[0.0211], s_loss:[0.0084] 979 | lr: 3.125000e-06, epoch: [92500/150000], all_loss: [0.0342], m_loss:[0.0165], s_loss:[0.0177] 980 | iCoseg Cosal2015 f: [0.8423], m: [0.0723], f: [0.8532], m: [0.0791] 981 | lr: 3.125000e-06, epoch: [92600/150000], all_loss: [0.0351], m_loss:[0.0209], s_loss:[0.0143] 982 | lr: 3.125000e-06, epoch: [92700/150000], all_loss: [0.0394], m_loss:[0.0284], s_loss:[0.0111] 983 | lr: 3.125000e-06, epoch: [92800/150000], all_loss: [0.0239], m_loss:[0.0152], s_loss:[0.0086] 984 | lr: 3.125000e-06, epoch: [92900/150000], all_loss: [0.0332], m_loss:[0.0252], s_loss:[0.0080] 985 | lr: 3.125000e-06, epoch: [93000/150000], all_loss: [0.0539], m_loss:[0.0327], s_loss:[0.0212] 986 | lr: 3.125000e-06, epoch: [93100/150000], all_loss: [0.0164], m_loss:[0.0109], s_loss:[0.0054] 987 | lr: 3.125000e-06, epoch: [93200/150000], all_loss: [0.0368], m_loss:[0.0244], s_loss:[0.0123] 988 | lr: 3.125000e-06, epoch: [93300/150000], all_loss: [0.0474], m_loss:[0.0302], s_loss:[0.0172] 989 | lr: 3.125000e-06, epoch: [93400/150000], all_loss: [0.0579], m_loss:[0.0404], s_loss:[0.0174] 990 | lr: 3.125000e-06, epoch: [93500/150000], all_loss: [0.0206], m_loss:[0.0133], s_loss:[0.0072] 991 | lr: 3.125000e-06, epoch: [93600/150000], all_loss: [0.0466], m_loss:[0.0326], s_loss:[0.0140] 992 | lr: 3.125000e-06, epoch: [93700/150000], all_loss: [0.0687], m_loss:[0.0528], s_loss:[0.0158] 993 | lr: 3.125000e-06, epoch: [93800/150000], all_loss: [0.0412], m_loss:[0.0282], s_loss:[0.0130] 994 | lr: 3.125000e-06, epoch: [93900/150000], all_loss: [0.0451], m_loss:[0.0312], s_loss:[0.0139] 995 | lr: 3.125000e-06, epoch: [94000/150000], all_loss: [0.0486], m_loss:[0.0337], s_loss:[0.0149] 996 | lr: 3.125000e-06, epoch: [94100/150000], all_loss: [0.0295], m_loss:[0.0180], s_loss:[0.0115] 997 | lr: 3.125000e-06, epoch: [94200/150000], all_loss: [0.0463], m_loss:[0.0287], s_loss:[0.0177] 998 | lr: 3.125000e-06, epoch: [94300/150000], all_loss: [0.0313], m_loss:[0.0217], s_loss:[0.0095] 999 | lr: 3.125000e-06, epoch: [94400/150000], all_loss: [0.0461], m_loss:[0.0316], s_loss:[0.0144] 1000 | lr: 3.125000e-06, epoch: [94500/150000], all_loss: [0.0307], m_loss:[0.0180], s_loss:[0.0127] 1001 | lr: 3.125000e-06, epoch: [94600/150000], all_loss: [0.0449], m_loss:[0.0292], s_loss:[0.0158] 1002 | lr: 3.125000e-06, epoch: [94700/150000], all_loss: [0.0476], m_loss:[0.0267], s_loss:[0.0209] 1003 | lr: 3.125000e-06, epoch: [94800/150000], all_loss: [0.0672], m_loss:[0.0465], s_loss:[0.0208] 1004 | lr: 3.125000e-06, epoch: [94900/150000], all_loss: [0.0610], m_loss:[0.0367], s_loss:[0.0243] 1005 | lr: 3.125000e-06, epoch: [95000/150000], all_loss: [0.0878], m_loss:[0.0672], s_loss:[0.0206] 1006 | iCoseg Cosal2015 f: [0.8407], m: [0.0730], f: [0.8531], m: [0.0788] 1007 | lr: 3.125000e-06, epoch: [95100/150000], all_loss: [0.0312], m_loss:[0.0226], s_loss:[0.0086] 1008 | lr: 3.125000e-06, epoch: [95200/150000], all_loss: [0.0541], m_loss:[0.0424], s_loss:[0.0117] 1009 | lr: 3.125000e-06, epoch: [95300/150000], all_loss: [0.0445], m_loss:[0.0245], s_loss:[0.0200] 1010 | lr: 3.125000e-06, epoch: [95400/150000], all_loss: [0.0359], m_loss:[0.0224], s_loss:[0.0135] 1011 | lr: 3.125000e-06, epoch: [95500/150000], all_loss: [0.0682], m_loss:[0.0448], s_loss:[0.0234] 1012 | lr: 3.125000e-06, epoch: [95600/150000], all_loss: [0.0386], m_loss:[0.0269], s_loss:[0.0116] 1013 | lr: 3.125000e-06, epoch: [95700/150000], all_loss: [0.0650], m_loss:[0.0471], s_loss:[0.0178] 1014 | --------------------------------------------------------------------------------