├── .gitignore └── README.md /.gitignore: -------------------------------------------------------------------------------- 1 | .DS_Store 2 | .idea 3 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | ## Real-time network for mobile devices 2 | 3 | ### Google 4 | 5 | 1. Learning Transferable Architectures for Scalable Image Recognition [NASNet](https://arxiv.org/abs/1707.07012) 6 | 2. MnasNet: Platform-Aware Neural Architecture Search for Mobile [MnasNet](https://arxiv.org/abs/1807.11626 ) 7 | 3. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications [MobileNets](https://arxiv.org/abs/1704.04861) 8 | 4. MobileNetV2: Inverted Residuals and Linear Bottlenecks [MobileNetV2](https://arxiv.org/pdf/1801.04381.pdf) [MobileNetV2-pytorch](https://github.com/Randl/MobileNetV2-pytorch) 9 | 5. MorphNet: Fast & Simple Resource-Constrained Structure Learning of Deep Networks [CVPR2018](http://openaccess.thecvf.com/content_cvpr_2018/papers/Gordon_MorphNet_Fast__CVPR_2018_paper.pdf) [code](https://github.com/google-research/morph-net) 10 | 6. Searching for MobileNetV3 [MobileNetV3](https://arxiv.org/pdf/1905.02244.pdf) **unofficial implementation** [MobileNetV3-for-Segmentation](https://github.com/Vipermdl/MobileNetV3-for-Segmentation) [official TF Repo](https://github.com/tensorflow/models/tree/master/research/object_detection) 11 | 7. Efficient Net: Rethinking Model Scaling for Convolutional Neural Networks. [ICML 2019](https://arxiv.org/pdf/1905.11946.pdf) [tfcode](https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet) [unofficial pytorch version](https://github.com/zsef123/EfficientNets-PyTorch) 12 | 8. MixNet: Mixed Depthwise Convolutional Kernels [BMVC2019](https://arxiv.org/pdf/1907.09595.pdf) [Official TF Repo](https://github.com/tensorflow/tpu/tree/master/models/official/mnasnet/mixnet) [Unofficial pytporch](https://github.com/zsef123/MixNet-PyTorch) 13 | 9. EfficientNetV2: Smaller Models and Faster Training [arxiv2021](https://arxiv.org/abs/2104.00298) [EfficientNetV2](https://github.com/google/automl/efficientnetv2) 14 | 15 | ### Microsoft Research 16 | 17 | 1. Interleaved Group Convolutions for Deep Neural Networks [IGCV](https://arxiv.org/abs/1707.02725) 18 | 2. IGCV2: Interleaved Structured Sparse Convolutional Neural Networks [IGCV2](https://arxiv.org/pdf/1804.06202.pdf) 19 | 3. IGCV3: Interleaved Low-Rank Group Convolutions for Efficient Deep Neural Networks [IGCV3](https://arxiv.org/abs/1806.00178) 20 | 4. Accelerating Deep Neural Networks with Spatial Bottleneck Modules [arxiv2018](https://arxiv.org/pdf/1809.02601.pdf) 21 | 5. Dynamic Convolution: Attention over Convolution Kernels [arxiv2019](https://arxiv.org/pdf/1912.03458.pdf) 22 | 23 | ### Facebook 24 | 25 | 1. Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution [arxiv2019](https://arxiv.org/pdf/1904.05049.pdf) **unofficial implementation** [OctaveConv_pytorch](https://github.com/lxtGH/OctaveConv_pytorch) [OctaveConv_MXNet](https://github.com/terrychenism/OctaveConv) 26 | 2. FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search [arxiv2018](https://arxiv.org/pdf/1812.03443.pdf) [code](https://github.com/facebookresearch/mobile-vision) 27 | 3. ChamNet: Towards Efficient Network Design through Platform-Aware Model Adaptation [arxiv2018](https://arxiv.org/pdf/1812.08934.pdf) [code](https://github.com/facebookresearch/mobile-vision) 28 | 4. Hybrid Composition with IdleBlock: More Efficient Networks for Image Recognition [arxiv2019](https://arxiv.org/pdf/1911.08609.pdf) 29 | 5. FBNetV2: Differentiable Neural Architecture Search for Spatial and Channel Dimensions [CVPR2020](https://128.84.21.199/pdf/2004.05565.pdf) [FBNetv2](https://github.com/facebookresearch/mobile-vision) 30 | 6. FBNetV3: Joint Architecture-Recipe Search using Neural Acquisition Function [arxiv2020](https://arxiv.org/pdf/2006.02049.pdf) 31 | 7. FP-NAS: Fast Probabilistic Neural Architecture Search [arxiv2020](https://arxiv.org/abs/2011.10949) 32 | 8. Fast and Accurate Model Scaling [arxiv2021](https://arxiv.org/abs/2103.06877) 33 | 9. Mobile Computer Vision @ Facebook [mobile-vision](https://github.com/facebookresearch/mobile-vision) 34 | 35 | 36 | ### Apple 37 | 1. CVNets: High Performance Library for Computer Vision [Technical report 4 Jun 2022](https://arxiv.org/abs/2206.02002) [ml-cvnets](https://github.com/apple/ml-cvnets) 38 | 39 | ### Megvill 40 | 41 | 1. ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices [ShuffleNet](https://arxiv.org/abs/1707.01083) 42 | 2. ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design [ShuffleNet V2]( https://arxiv.org/abs/1807.11164) [Shufflenet-v2-Pytorch](https://github.com/ericsun99/Shufflenet-v2-Pytorch) 43 | 3. ShuffleNetV2+：[paper unrelease]() [code](https://github.com/megvii-model/ShuffleNet-Series/tree/master/ShuffleNetV2%2B) *ShuffleNet Series by Megvii Research* 44 | 4. WeightNet: Revisiting the Design Space of Weight Networks [ECCV2020](https://arxiv.org/pdf/2007.11823.pdf) [WeightNet](https://github.com/megvii-model/WeightNet) 45 | 5. RepVGG: Making VGG-style ConvNets Great Again [arxiv2020](https://arxiv.org/abs/2101.03697) [RepVGG](https://github.com/DingXiaoH/RepVGG) *Tsinghua University, MEGVII Technology, etc* 46 | 47 | ### Others 48 | 49 | 1. CondenseNet: An Efficient DenseNet using Learned Group Convolutions [CondenseNet](https://arxiv.org/abs/1711.09224) 50 | 2. CondenseNet V2: Sparse Feature Reactivation for Deep Networks [CondenseNetV2](https://github.com/jianghaojun/CondenseNetV2) 51 | 3. ANTNets: Mobile Convolutional Neural Networks for Resource Efficient Image Classification [arxiv2019](https://arxiv.org/pdf/1904.03775.pdf) 52 | 4. Seesaw-Net: Convolution Neural Network With Uneven Group Convolution [arxiv2019](https://arxiv.org/pdf/1905.03672.pdf) 53 | 5. ISBNet: Instance-aware Selective Branching Network [arxiv2019](https://arxiv.org/pdf/1905.04849.pdf) 54 | 6. Multinomial Distribution Learning for Effective Neural Architecture Search [arxiv2019](https://arxiv.org/pdf/1905.07529.pdf) [code](https://github.com/tanglang96/MDENAS) 55 | 7. HGC: Hierarchical Group Convolution for Highcdly Efficient Neural Network [arxiv2019](https://128.84.21.199/pdf/1906.03657.pdf) 56 | 8. DiCENet: Dimension-wise Convolutions for Efficient Networks [arxiv2019](https://arxiv.org/pdf/1906.03516.pdf) [code](https://github.com/sacmehta/EdgeNets) 57 | 9. Densely Connected Search Space for More Flexible Neural Architecture Search [arxiv2019](https://arxiv.org/pdf/1906.09607.pdf) [code](https://github.com/JaminFong/DenseNAS) 58 | 10. VarGNet: Variable Group Convolutional Neural Network for Efficient Embedded Computing [arxiv2019](https://arxiv.org/pdf/1907.05653.pdf) *Horizon Robotics* 59 | 11. VarGFaceNet: An Efficient Variable Group Convolutional Neural Network for Lightweight Face Recognition [VarGFaceNet](http://openaccess.thecvf.com/content_ICCVW_2019/papers/LSR/Yan_VarGFaceNet_An_Efficient_Variable_Group_Convolutional_Neural_Network_for_Lightweight_ICCVW_2019_paper.pdf) [code](https://github.com/zma-c-137/VarGFaceNet) *Horizon Robotics* (1st place in The Lightweight Face Recognition Challenge & Workshop ICCV 2019) 60 | 12. Once for All: Train One Network and Specialize it for Efficient Deployment [ICLR 2020](https://arxiv.org/pdf/1908.09791.pdf) [ofaNet](https://github.com/mit-han-lab/once-for-all) *MIT && MIT-IBM Watson AI Lab* 61 | 13. LPRNet: Lightweight Deep Network by Low-rank Pointwise Residual Convolution [arxiv2019](https://arxiv.org/pdf/1910.11853.pdf) 62 | 14. LeanConvNets: Low-cost Yet Effective Convolutional Neural Networks [arxiv2019](https://arxiv.org/pdf/1910.13157v1.pdf) *Ben-Gurion University&&Emory Universit* 63 | 15. DeepSquare: Boosting the Learning Power of Deep Convolutional Neural Networks with Elementwise Square Operators [arxiv2019](https://arxiv.org/pdf/1906.04979.pdf) *ByteDance AI Lab* 64 | 16. ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks [arxiv2019](https://arxiv.org/pdf/1910.03151.pdf) [code](https://github.com/BangguWu/ECANet) *Tianjin University* 65 | 17. XSepConv: Extremely Separated Convolution [arxiv2020](https://arxiv.org/pdf/2002.12046.pdf) *Tsinghua University &&University College London* 66 | 18. GhostNet: More Features from Cheap Operations [CVPR2020 ](https://arxiv.org/abs/1911.11907) [TF Repo](https://github.com/huawei-noah/ghostnet) *Huawei Noah’s Ark Lab&&Peking University&&Sydney University* 67 | 19. Resolution Adaptive Networks for Efficient Inference [CVPR2020](https://arxiv.org/pdf/2003.07326.pdf) [RANet](https://github.com/yangle15/RANet-pytorch/blob/master/models/RANet.py) *Tsinghua && HIT &&SenseTime* 68 | 20. TResNet: High Performance GPU-Dedicated Architecture [arxiv2020](https://arxiv.org/pdf/2003.13630.pdf) [TResNet](https://github.com/mrT23/TResNet) *DAMO Academy, Alibaba Group* 69 | 21. Neural Architecture Design for GPU-Efficient Networks [arxiv2020](https://arxiv.org/pdf/2006.14090.pdf) *Alibaba Group* 70 | 22. ReXNet: Diminishing Representational Bottleneck on Convolutional Neural Network [arxiv2020](https://arxiv.org/pdf/2007.00992) [rexnet](https://github.com/clovaai/rexnet) *Clova AI Research, NAVER Corp* 71 | 23. Rethinking Bottleneck Structure for Efficient Mobile Network Design [ECCV2020](https://arxiv.org/pdf/2007.02269) [MobileNeXt](https://github.com/RangiLyu/mobilenext) [yitu-opensource](https://github.com/yitu-opensource/MobileNeXt) 72 | 24. Model Rubik’s Cube: Twisting Resolution, Depth and Width for TinyNets [NeurIPS 2020](https://arxiv.org/abs/2010.14819)** 73 | 25. MicroNet: Towards Image Recognition with Extremely Low FLOPs [arxiv2020](https://arxiv.org/abs/2011.12289v1) *UC San Diego && Microsoft* 74 | 26. Lite-HRNet: A Lightweight High-Resolution Network [cvpr2021](https://arxiv.org/abs/2104.06403) [Lite-HRNet](https://github.com/HRNet/Lite-HRNet) 75 | 27. AsymmNet: Towards ultralight convolution neural networks using asymmetrical bottlenecks [MAI@CVPR 2021](https://arxiv.org/abs/2104.07770) [AsymmNet](https://github.com/Spark001/AsymmNet) 76 | 77 | ### Semantic segmentation 78 | 79 | 1. ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation [ENet](https://arxiv.org/abs/1606.02147) 80 | 2. ICNet for Real-Time Semantic Segmentation on High-Resolution Images [ICNet](https://arxiv.org/abs/1704.08545) 81 | 3. Speeding up Semantic Segmentation for Autonomous Driving [SQNet](https://openreview.net/pdf?id=S1uHiFyyg) 82 | 4. ERFNet: Efficient Residual Factorized ConvNet for Real-time Semantic Segmentation [ERFNet](http://www.robesafe.es/personal/eduardo.romera/pdfs/Romera17tits.pdf) 83 | 5. ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation [ESPNet](https://arxiv.org/abs/1803.06815) 84 | 6. BiSeNet: Bilateral Segmentation Network for Real-time Semantic Segmentation [ECCV2018](https://arxiv.org/abs/1808.00897) [code](https://github.com/ycszen/TorchSeg) 85 | 7. [A Comparative Study of Real-time Semantic Segmentation for Autonomous Driving](http://openaccess.thecvf.com/content_cvpr_2018_workshops/papers/w12/Siam_A_Comparative_Study_CVPR_2018_paper.pdf) 86 | 8. Efficient Dense Modules of Asymmetric Convolution for Real-Time Semantic Segmentation [EDANet](https://github.com/shaoyuanlo/EDANet) 87 | 9. Light-Weight RefineNet for Real-Time Semantic Segmentation [Light-Weight RefineNet](https://github.com/drsleep/light-weight-refinenet) 88 | 10. [Searching for Efficient Multi-Scale Architectures for Dense Image Prediction](https://github.com/tensorflow/models/tree/master/research/deeplab) 89 | 11. CGNet: A Light-weight Context Guided Network for Semantic Segmentation [arxiv2018](https://arxiv.org/abs/1811.08201) [code](https://github.com/wutianyiRosun/CGNet) 90 | 12. ShelfNet for Real-time Semantic Segmentation [arxiv2018](https://arxiv.org/pdf/1811.11254v1.pdf) [code](https://github.com/juntang-zhuang/ShelfNet) 91 | 13. Concentrated-Comprehensive Convolutions for lightweight semantic segmentation [arxiv2018](https://arxiv.org/pdf/1812.04920.pdf) 92 | 14. ESPNetv2: A Light-weight, Power Efficient, and General Purpose Convolutional Neural Network [cvpr2019](https://arxiv.org/pdf/1811.11431.pdf) [code](https://github.com/sacmehta/ESPNetv2) 93 | 15. Fast-SCNN: Fast Semantic Segmentation Network [arxiv2019](https://arxiv.org/pdf/1902.04502.pdf) [code](https://github.com/kshitizrimal/Fast-SCNN/blob/master/tf_2_0_fast_scnn.py) [blog](https://medium.com/deep-learning-journals/fast-scnn-explained-and-implemented-using-tensorflow-2-0-6bd17c17a49e) 94 | 16. An efficient solution for semantic segmentation: ShuffleNet V2 with atrous separable convolutions [arxiv2019](https://arxiv.org/pdf/1902.07476.pdf) 95 | 17. Decoders Matter for Semantic Segmentation: Data-Dependent Decoding Enables Flexible Feature Aggregation [cvpr2019](https://arxiv.org/pdf/1903.02120.pdf) 96 | 18. DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation [cvpr2019Megvii ](https://arxiv.org/pdf/1904.02216.pdf) 97 | 19. Real time backbone for semantic segmentation [arxiv2019](https://arxiv.org/pdf/1903.06922.pdf) 98 | 20. Residual Pyramid Learning for Single-Shot Semantic Segmentation [arxiv2019](https://128.84.21.199/pdf/1903.09746.pdf) 99 | 21. Knowledge Adaptation for Efficient Semantic Segmentation [cvpr2019](https://arxiv.org/pdf/1903.04688.pdf) 100 | 22. In Defense of Pre-trained ImageNet Architectures for Real-time Semantic Segmentation of Road-driving Images [cvpr2019](https://arxiv.org/pdf/1903.08469.pdf) [code](https://github.com/orsic/swiftnet) 101 | 23. Towards Real-Time Automatic Portrait Matting on Mobile Devices [arxiv2019](https://arxiv.org/pdf/1904.03816.pdf) [code](https://github.com/hyperconnect/MMNet) 102 | 24. PortraitNet: Real-time Portrait Segmentation Network for Mobile Device [Computers & Graphics 2019](http://www.yongliangyang.net/docs/mobilePotrait_c&g19.pdf) [code](https://github.com/wpf535236337/pytorch_PortraitNet) 103 | 25. Design of Real-time Semantic Segmentation Decoder for Automated Driving [VISAPP2019](https://arxiv.org/pdf/1901.06580.pdf) 104 | 26. ThunderNet: A Turbo Unified Network for Real-Time Semantic Segmentation [WACV2019](http://vlm1.uta.edu/~athitsos/publications/xiang_wacv2019.pdf) 105 | 27. LEDNet: A Lightweight Encoder-Decoder Network for Real-time Semantic Segmentation [ICIP2019](https://arxiv.org/pdf/1905.02423.pdf) [code](https://github.com/xiaoyufenfei/LEDNet) 106 | 28. Accurate Facial Image Parsing at Real-Time Speed [TIP2019](http://www.colalab.org/media/paper/bare_jrnl_EjnyBQb.pdf) 107 | 29. Efficient Ladder-style DenseNets for Semantic Segmentation of Large Images [arxiv2019](https://arxiv.org/pdf/1905.05661.pdf) 108 | 30. Nail Polish Try-On: Realtime Semantic Segmentation of Small Objects forNative and Browser Smartphone AR Applications [ CVPRW2019](https://arxiv.org/pdf/1906.02222v1.pdf) 109 | 31. ESNet: An Efficient Symmetric Network for Real-time Semantic Segmentation [arxiv2019](https://arxiv.org/pdf/1906.09826.pdf) [code](https://github.com/xiaoyufenfei/ESNet) 110 | 32. Real-time Hair Segmentation and Recoloring on Mobile GPUs [2019CVPRW](https://arxiv.org/pdf/1907.06740.pdf) 111 | 33. Efficient Segmentation: Learning Downsampling Near Semantic Boundaries [arxiv2019](https://arxiv.org/pdf/1907.07156v1.pdf) 112 | 34. Multi-Class Lane Semantic Segmentation using Efficient Convolutional Networks [MMSP 2019](https://arxiv.org/pdf/1907.09438.pdf) 113 | 35. A Comparative Study of High-Recall Real-Time Semantic Segmentation Based on Swift Factorized Network [SPIE Security + Defence 2019](https://arxiv.org/pdf/1907.11394.pdf) [code](https://github.com/Katexiang/swiftnet/tree/master/Swift_Factorized_Network(SFN)) 114 | 36. Context-Integrated and Feature-Refined Network for Lightweight Urban Scene Parsing [arxiv2019](https://arxiv.org/pdf/1907.11474.pdf) 115 | 37. DABNet: Depth-wise Asymmetric Bottleneck for Real-time Semantic Segmentation [BMVC2019](https://arxiv.org/pdf/1907.11357v2.pdf) [code](https://github.com/Reagan1311/DABNet) 116 | 38. Learning Lightweight Lane Detection CNNs by Self Attention Distillation [ICCV2019](https://arxiv.org/pdf/1908.00821.pdf) [torchRepo](https://github.com/cardwing/Codes-for-Lane-Detection) 117 | 39. SqueezeNAS: Fast neural architecture search for faster semantic segmentation [arxiv2019](https://arxiv.org/pdf/1908.01748.pdf) [code](https://github.com/ashaw596/squeezenas) *submitted to ICCV Neural Architects workshop* 118 | 40. ExtremeC3Net: Extreme Lightweight Portrait Segmentation Networks using Advanced C3-modules [arxiv2019](https://arxiv.org/pdf/1908.03093v1.pdf) [ExtremeC3Net](https://github.com/clovaai/ext_portrait_segmentation) *Seoul National University && Clova AI, NAVER Corp* 119 | 41. Customizable Architecture Search for Semantic Segmentation [cvpr2019](http://openaccess.thecvf.com/content_CVPR_2019/papers/Zhang_Customizable_Architecture_Search_for_Semantic_Segmentation_CVPR_2019_paper.pdf) *University of Science and Technology of China && JD AI Research* 120 | 42. See More than Once – Kernel-Sharing Atrous Convolution for Semantic Segmentation [arxiv2019](https://arxiv.org/pdf/1908.09443.pdf) *University of Technology Sydney && East China Normal University* 121 | 43. Feature Pyramid Encoding Network for Real-time Semantic Segmentation [BMVC2019](https://arxiv.org/pdf/1909.08599.pdf) *The University of Manchester* 122 | 44. Graph-guided Architecture Search for Real-time Semantic Segmentation [arxiv2019](https://arxiv.org/pdf/1909.06793.pdf) *SenseTime Research && Zhejiang University* 123 | 45. Eye Semantic Segmentation with a Lightweight Model [ICCVW 2019]( https://arxiv.org/pdf/1911.01049.pdf) [code](https://github.com/th2l/Eye_VR_Segmentation) *Chonnam National University* 124 | 46. FDDWNet: A Lightweight Convolutional Neural Network for Real-time Sementic Segmentation [arxiv2019](https://arxiv.org/pdf/1911.00632.pdf) *Nanjing University of Posts & Telecommunications* 125 | 47. Real-Time Semantic Segmentation via Multiply Spatial Fusion Network [arxiv2019](https://arxiv.org/pdf/1911.07217.pdf) *Beihang University &&Megvii Technology&&Peng Cheng Laboratory* 126 | 48. RGPNet: A Real-Time General Purpose Semantic Segmentation [arxiv2019](https://arxiv.org/pdf/1912.01394.pdf) *Advanced Research Lab, NavInfo Europe* 127 | 49. LiteSeg: A Novel Lightweight ConvNet for Semantic Segmentation [DICTA2019](https://arxiv.org/pdf/1912.06683.pdf) 128 | 50. FasterSeg: Searching for Faster Real-time Semantic Segmentation [ICIR2020](https://openreview.net/pdf?id=BJgqQ6NYvB) [FasterSeg](https://github.com/TAMU-VITA/FasterSeg) *Texas A&M University && Horizon Robotics Inc* 129 | 51. Real-time Segmentation and Facial Skin Tones Grading [arxiv2019](https://arxiv.org/pdf/1912.12888.pdf) [HLRNet](https://github.com/JACKYLUO1991/Face-skin-hair-segmentaiton-and-skin-color-evaluation) 130 | 52. SINet: Extreme Lightweight Portrait Segmentation Networks with Spatial Squeeze Modules and Information Blocking Decoder [WACV2020](https://arxiv.org/pdf/1911.09099.pdf) [SINet](https://github.com/clovaai/ext_portrait_segmentation) 131 | 53. FarSee-Net: Real-Time Semantic Segmentation by Efficient Multi-scale Context Aggregation and Feature Space Super-resolution [ICRA2020](https://128.84.21.199/pdf/2003.03913.pdf) *SenseTime && Tokyo University* 132 | 54. Cars Can’t Fly up in the Sky: Improving Urban-Scene Segmentation via Height-driven Attention Networks [CVPR2020](https://128.84.21.199/pdf/2003.05128.pdf) [HANet](https://github.com/shachoi/HANet) *Korea University && CA* 133 | 55. Real-Time High-Performance Semantic Image Segmentation of Urban Street Scenes [arxiv2020](https://arxiv.org/pdf/2003.08736.pdf) *XMU* 134 | 56. Real-Time Semantic Segmentation via Auto Depth, Downsampling Joint Decision and Feature Aggregation [arxiv2020](https://arxiv.org/pdf/2003.14226.pdf) *ZheJiang University* 135 | 57. BiSeNet V2: Bilateral Network with Guided Aggregation for Real-time Semantic Segmentation [arxiv2020](https://arxiv.org/pdf/2004.02147.pdf) [BiSeNet V2](https://github.com/ycszen/BiSeNet) [bisenetv2-tensorflow](https://github.com/MaybeShewill-CV/bisenetv2-tensorflow) *HUST && The University of Adelaide &&CUHK && Tencent* 136 | 58. Bi-direction Context Propagation Network for Real-time Semantic Segmentation [arxiv2020](https://arxiv.org/pdf/2005.11034.pdf) *Hefei University of Technology* 137 | 59. LRNNet: A Light-Weighted Network with Efficient Reduced Non-Local Operation for Real-Time Semantic Segmentation [arxiv2020](https://arxiv.org/pdf/2006.02706v1.pdf) *SJTU* 138 | 60. Real-time Semantic Segmentation with Fast Attention [arxiv2020](https://arxiv.org/pdf/2007.03815.pdf) [Homepage](https://cs-people.bu.edu/pinghu/FANet.html) [FANet](https://github.com/feinanshan/FANet) *Boston University && Adobe Research* 139 | 61. Ultra Fast Structure-aware Deep Lane Detection [ECCV2020](https://arxiv.org/pdf/2007.03815.pdf) [code](https://github.com/cfzd/Ultra-Fast-Lane-Detection) *Zhejiang University* 140 | 62. Improving Semantic Segmentation via Decoupled Body and Edge Supervision [ECCV2020](https://arxiv.org/pdf/2007.10035v1.pdf) [DecoupleSegNets](https://github.com/lxtGH/DecoupleSegNets) *Peking University && University of Oxford && SenseTime Research && Zhejiang Lab* 141 | 63. Semantic Segmentation With Multi Scale Spatial Attention For Self Driving Cars [arxiv2020](https://arxiv.org/pdf/2007.12685.pdf) 142 | 64. EfficientSeg: An Efficient Semantic Segmentation Network [arxiv2020](https://arxiv.org/pdf/2009.06469.pdf) [EfficientSeg](https://github.com/MrGranddy/EfficientSeg) 143 | 65. Dense Dual-Path Network for Real-time Semantic Segmentation [ACCV2020](https://arxiv.org/abs/2010.10778) 144 | 66. CABiNet: Efficient Context Aggregation Network for Low-Latency Semantic Segmentation [arxiv2020](https://arxiv.org/abs/2011.00993) 145 | 67. SegBlocks: Block-Based Dynamic Resolution Networks for Real-Time Segmentation [arxiv2020](https://arxiv.org/abs/2011.12025) *KU Leuven* 146 | 68. Real-Time High-Resolution Background Matting [arxiv2020](https://arxiv.org/abs/2012.07810) [hompage](https://grail.cs.washington.edu/projects/background-matting-v2/) [BackgroundMattingV2](https://github.com/PeterL1n/BackgroundMattingV2) *University of Washington* 147 | 69. Is a Green Screen Really Necessary for Real-Time Portrait Matting? [arxiv2020](https://arxiv.org/abs/2011.11961) [MODNet](https://github.com/ZHKKKe/MODNet) *City University of Hong Kong, SenseTime Research* 148 | 70. HyperSeg: Patch-wise Hypernetwork for Real-time Semantic Segmentation [arxiv2020](https://arxiv.org/abs/2012.11582) *Facebook AI, Tel Aviv University* 149 | 71. Boundary-Aware Segmentation Network for Mobile and Web Applications [submitted to TPAMI](https://arxiv.org/abs/2101.04704) [BASNet](https://github.com/NathanUA/BASNet) 150 | 72. EADNet: Efficient Asymmetric Dilated Network for Semantic Segmentation [arxiv2021](https://arxiv.org/abs/2103.08914) *Fudan University , etc.* 151 | 73. AttaNet: Attention-Augmented Network for Fast and Accurate Scene Parsing [AAAI2021](https://arxiv.org/abs/2103.05930) 152 | 74. Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes [arxiv2021](https://arxiv.org/abs/2101.06085) 153 | 75. Rethinking BiSeNet For Real-time Semantic Segmentation [CVPR2021](https://arxiv.org/abs/2104.13188) [STDC-Seg](https://github.com/MichaelFan01/STDC-Seg) 154 | 76. TopFormer: Token Pyramid Transformer for Mobile Semantic Segmentation [CVPR2022](https://arxiv.org/abs/2204.05525) [code](https://github.com/hustvl/TopFormer) 155 | 77. 156 | 157 | ### Instance Segmentation 158 | 159 | 1. YOLACT: Real-time Instance Segmentation [ICCV2019](https://arxiv.org/pdf/1904.02689.pdf) [code](https://github.com/dbolya/yolact) *University of California, Davis* 160 | 2. CenterMask:Real-Time Anchor-Free Instance Segmentation [VovnetV2](https://arxiv.org/pdf/1911.06667.pdf) [CenterMask](https://github.com/youngwanLEE/CenterMask) [CenterMask2](https://github.com/youngwanLEE/centermask2) *ETRI* 161 | 3. YOLACT++ Better Real-time Instance Segmentation [arxiv2019](YOLACT++ Better Real-time Instance Segmentation) *University of California, Davis* 162 | 4. BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation [arxiv2020](https://arxiv.org/pdf/2001.00309.pdf) *The University of Adelaide && Southeast University &&Huawei Noah’s Ark Lab* 163 | 5. Deep Snake for Real-Time Instance Segmentation [cvpr2020](https://arxiv.org/pdf/2001.01629.pdf) [snake](https://github.com/zju3dv/snake) *Zhejiang University* 164 | 6. PolarMask: Single Shot Instance Segmentation with Polar Representation [cvpr2020](https://arxiv.org/pdf/1909.13226.pdf) [PolarMask](https://github.com/xieenze/PolarMask) *HKU* 165 | 7. SOLOv2: Dynamic, Faster and Stronger [arxiv2020](https://arxiv.org/pdf/2003.10152.pdf) [SOLOV2](https://github.com/aim-uofa/AdelaiDet) [UnofficialPytorchRepo](https://github.com/Epiphqny/SOLOv2) *The University of Adelaide && Tongji University && ByteDance AI Lab* 166 | 8. YolactEdge: Real-time Instance Segmentation on the Edge [arxiv2020](https://arxiv.org/abs/2012.12259) [YolactEdge](https://github.com/haotian-liu/yolact_edge) 167 | 168 | ### Salient Object Detection 169 | 170 | * U2-Net: Going Deeper with Nested U-Structure for Salient Object Detection [PR2020](https://arxiv.org/abs/2005.09007) [U-2-Net](https://github.com/NathanUA/U-2-Net) 171 | * Highly Efficient Salient Object Detection with 100K Parameters [ECCV2020](https://arxiv.org/abs/2003.05643) [CSNet](https://github.com/MCG-NKU/SOD100K) 172 | * MobileSal: Extremely Efficient RGB-D Salient Object Detection [arxiv2020](https://arxiv.org/abs/2012.13095) 173 | 174 | ### Object detection 175 | 176 | 1. An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection [CVPR2019 CEFRL Workshop](https://arxiv.org/pdf/1904.09730.pdf) [VOVNetPytorch](https://github.com/stigma0617/VoVNet.pytorch) *ETRI* 177 | 2. EfficientDet: Scalable and Efficient Object Detection [arxiv2019](https://arxiv.org/pdf/1911.09070.pdf) [unofficial-EfficientDet.Pytorch](https://github.com/toandaominh1997/EfficientDet.Pytorch) *Google Research, Brain Team* 178 | 2. RefineDetLite: A Lightweight One-stage Object Detection Framework for CPU-only Devices [arxiv2019](https://arxiv.org/pdf/1911.08855.pdf) *Tencent Research&&HKUST* 179 | 3. Learning Spatial Fusion for Single-Shot Object Detection [arxiv2019](https://arxiv.org/pdf/1911.09516) [yolov3+](https://github.com/ruinmessi/ASFF) *Beihang University* 180 | 5. CSPNet: A New Backbone that can Enhance Learning Capability of CNN [arxiv2019](https://arxiv.org/pdf/1911.11929.pdf) [CSPNet](https://github.com/WongKinYiu/CrossStagePartialNetworks) *research teams from TaiWan* 181 | 6. YOLOv4: Optimal Speed and Accuracy of Object Detection [arxiv2020](https://arxiv.org/pdf/2004.10934.pdf) [OfficialRepo](https://github.com/AlexeyAB/darknet) [Minimal PyTorch](https://github.com/Tianxiaomo/pytorch-YOLOv4) [Tensorflow 2.0Repo](https://github.com/hunglc007/tensorflow-yolov4-tflite) [KerasRepo](https://github.com/Ma-Dan/keras-yolo4) 182 | 7. YOLO Nano: a Highly Compact You Only Look Once Convolutional Neural Network for Object Detection [arxiv2019](https://arxiv.org/abs/1910.01271) 183 | 184 | ### Neural Architecture Search 185 | 186 | 1. PC-DARTS: Partial Channel Connections for Memory-Efficient Differentiable Architecture Search [arxiv2019](https://arxiv.org/pdf/1907.05737.pdf) [code](https://github.com/yuhuixu1993/PC-DARTS) *Shanghai Jiao Tong University&&Huawei* 187 | 2. Densely Connected Search Space for More Flexible Neural Architecture Search [arxiv2019](https://arxiv.org/pdf/1906.09607.pdf) [code](https://github.com/JaminFong/DenseNAS) *Huazhong University of Science and Technology &&Horizon Robotics* 188 | 3. FairNAS: Rethinking Evaluation Fairness of Weight Sharing Neural Architecture Search [arxiv2019](https://arxiv.org/pdf/1907.01845.pdf) [code](https://github.com/fairnas/FairNAS) *Xiaomi AI Lab* 189 | 4. XferNAS: Transfer Neural Architecture Search [arxiv2019](https://arxiv.org/pdf/1907.08307.pdf) *IBM Research* 190 | 5. AutoML: A Survey of the State-of-the-Art [arxiv2019](https://arxiv.org/pdf/1908.00709.pdf) *Hong Kong Baptist University* 191 | 6. MoGA: Searching Beyond MobileNetV3 [arxiv2019](https://arxiv.org/abs/1908.01314) [code](https://github.com/xiaomi-automl/MoGA) *Xiaomi AI Lab* 192 | 7. ScarletNAS: Bridging the Gap Between Scalability and Fairness in Neural Architecture Search [arxiv2019](https://arxiv.org/pdf/1908.06022.pdf) [code](https://github.com/xiaomi-automl/SCARLET-NAS) *Xiaomi AI Lab && IoT* 193 | 8. BANANAS: Bayesian Optimization with Neural Architectures for Neural Architecture Search [arxiv](https://arxiv.org/pdf/1910.11858.pdf) [code](https://github.com/naszilla/bananas) [blog](https://medium.com/reality-engines/bananas-a-new-method-for-neural-architecture-search-192d21959c0c) 194 | 9. Fast and Practical Neural Architecture Search [iccv2019](http://jiaya.me/papers/fpnas_iccv19.pdf) *CUHK && YouTu Lab, Tencent* 195 | 10. Fair DARTS: Eliminating Unfair Advantages in Differentiable Architecture Search [arxiv2019](https://arxiv.org/pdf/1911.12126.pdf) [FairDARTS](https://github.com/xiaomi-automl/FairDARTS) *Xiaomi AI Lab &&Minzu University of China* 196 | 11. SGAS: Sequential Greedy Architecture Search [arxiv2019](https://arxiv.org/pdf/1912.00195.pdf) [project]() *KAUST && Intel Labs* 197 | 12. Blockwisely Supervised Neural Architecture Search with Knowledge Distillation [arxiv2019](https://arxiv.org/abs/1911.13053) [DNA](https://github.com/jiefengpeng/DNA) *DarkMatter AI Research && Monash University &&Sun Yat-sen University* 198 | 13. EDAS: Efficient and Differentiable Architecture Search [arxiv2019](https://128.84.21.199/pdf/1912.01237.pdf) *KAIST* 199 | 14. Efficient Differentiable Neural Architecture Search with Meta Kernels [arxiv2019](https://arxiv.org/pdf/1912.04749.pdf) *HUST &&YITU &&NUS* 200 | 15. AtomNAS: Fine-Grained End-to-End Neural Architecture Search [ICIR2020](https://openreview.net/pdf?id=BylQSxHFwr) [AutoNAS](https://github.com/meijieru/AtomNAS) *Johns Hopkins University && ByteDance AI Lab* 201 | 16. EcoNAS: Finding Proxies for Economical Neural Architecture Search [arxiv2020](https://arxiv.org/pdf/2001.01233.pdf) *The University of Sydney &&Nanyang Technological University&&SenseTime Research* 202 | 17. MixPath: A Unified Approach for One-shot Neural Architecture Search [arxiv2020](https://arxiv.org/pdf/2001.05887.pdf) [MixPath](https://github.com/xiaomi-automl/MixPath) *Xiaomi AI Lab &&UCAS* 203 | 204 | ### Transformer 205 | 206 | 1. Efficient Transformers: A Survey [arxiv2020](https://arxiv.org/abs/2009.06732v1) 207 | 2. Escaping the Big Data Paradigm with Compact Transformers [arxiv2021](https://arxiv.org/pdf/2104.05704.pdf) [compact- transformer](https://arxiv.org/pdf/2104.05704.pdf) 208 | 3. EdgeViTs: Competing Light-weight CNNs on Mobile Devices with Vision Transformers [arxiv2022](https://arxiv.org/abs/2205.03436) 209 | 4. MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer [ICLR2022](https://arxiv.org/abs/2110.02178)[code](https://github.com/apple/ml-cvnets) 210 | 5. EdgeFormer: Improving Light-weight ConvNets by Learning from Vision Transformers [arxiv2022](https://arxiv.org/abs/2203.03952) [code](https://github.com/hkzhang91/EdgeFormer) 211 | 6. SepViT: Separable Vision Transformer [arxiv2022](https://arxiv.org/abs/2203.15380) 212 | 7. MoCoViT: Mobile Convolutional Vision Transformer [arxiv2022](https://arxiv.org/abs/2205.12635) 213 | 8. Lightweight Vision Transformer with Cross Feature Attention[Technical Report](https://arxiv.org/abs/2207.07268) 214 | 9. Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios [arxiv2022](https://arxiv.org/abs/2207.05501) 215 | 216 | 217 | #### Tricks for training for Transformer 218 | 219 | 1. Training data-efficient image transformers & distillation through attention [arxiv2021](https://arxiv.org/abs/2012.12877) [DEIT](https://github.com/facebookresearch/deit) 220 | 2. Token Labeling: Training a 85.5% Top-1 Accuracy Vision Transformer 221 | with 56M Parameters on ImageNet [arxiv2021](https://arxiv.org/abs/2104.10858) [TokenLabeling](https://github.com/zihangJiang/TokenLabeling) 222 | 3. Improve Vision Transformers Training by Suppressing Over-smoothing [arxiv2021](https://arxiv.org/abs/2104.12753) [PatchVisionTransformer](https://github.com/ChengyueGongR/PatchVisionTransformer) 223 | 4. 224 | 225 | ### Tricks for training neural network 226 | 227 | 1. [A Recipe for Training Neural Networks](https://karpathy.github.io/2019/04/25/recipe/) (Apr 25, 2019) 228 | 2. [Improving deep learning models with bag of tricks](https://github.com/kmkolasinski/deep-learning-notes/tree/master/seminars/2018-12-Improving-DL-with-tricks) (Dec 13,2018) 229 | 3. [A Bag of Tricks for Image Classification](https://towardsdatascience.com/a-big-of-tricks-for-image-classification-fec41eb28e01) (Dec 17, 2018) 230 | 4. Bag of Tricks for Image Classification with Convolutional Neural Networks [cvpr2019](https://arxiv.org/abs/1812.01187) [code](https://github.com/dmlc/gluon-cv) 231 | 5. Bag of Freebies for Training Object Detection Neural Networks [arxiv2019](https://arxiv.org/abs/1902.04103) [code](https://github.com/dmlc/gluon-cv) 232 | 6. Bag of Tricks for Image Classification by Arthur Kuzin [2020slide](https://docs.google.com/presentation/d/1TcHW6aH0QrcLG_SazEjGx5RCsNjF2SQ9gzIR4RKOsRU/edit#slide=id.p) 233 | 7. carrier-of-tricks-for-classification-pytorch [2020code](https://github.com/hoya012/carrier-of-tricks-for-classification-pytorch) 234 | 8. [Faster Deep Learning Training with PyTorch – a 2021 Guide](https://efficientdl.com/faster-deep-learning-in-pytorch-a-guide/#17-use-input-and-batch-normalization) 235 | 236 | ### Some useful links 237 | 238 | 1. [Awesome-model-compression-and-acceleration](https://github.com/memoiry/Awesome-model-compression-and-acceleration) 239 | 2. [awesome-model-compression-and-acceleration](https://github.com/sun254/awesome-model-compression-and-acceleration) 240 | 3. [Model-Compression-Papers](https://github.com/chester256/Model-Compression-Papers) 241 | 4. [Awesome-model-compression-and-acceleration](https://github.com/memoiry/Awesome-model-compression-and-acceleration) 242 | 5. [awesome-AutoML-and-Lightweight-Models](https://github.com/guan-yuan/awesome-AutoML-and-Lightweight-Models) 243 | 6. [常用的语义分割架构结构综述以及代码复现](https://github.com/guanfuchen/semseg) 244 | 7. [Efficient-Segmentation-Networks](https://github.com/xiaoyufenfei/Efficient-Segmentation-Networks) 245 | 246 | 247 | 248 | ### Some blogs 249 | 250 | 1. [Real-time Portrait Segmentation on Smartphones](https://blog.prismalabs.ai/real-time-portrait-segmentation-on-smartphones-39c84f1b9e66) 251 | 2. [Mobile Real-time Video Segmentation](https://ai.googleblog.com/2018/03/mobile-real-time-video-segmentation.html) 252 | 3. [Real-Time deep learning in mobile application](https://medium.com/vitalify-asia/real-time-deep-learning-in-mobile-application-25cf601a8976) 253 | 4. [QNNPACK: Open source library for optimized mobile deep learning](https://code.fb.com/ml-applications/qnnpack/) 254 | 5. [第十七章 模型压缩及移动端部署](https://github.com/scutan90/DeepLearning-500-questions/blob/master/ch17_%E6%A8%A1%E5%9E%8B%E5%8E%8B%E7%BC%A9%E3%80%81%E5%8A%A0%E9%80%9F%E5%8F%8A%E7%A7%BB%E5%8A%A8%E7%AB%AF%E9%83%A8%E7%BD%B2/%E7%AC%AC%E5%8D%81%E4%B8%83%E7%AB%A0_%E6%A8%A1%E5%9E%8B%E5%8E%8B%E7%BC%A9%E3%80%81%E5%8A%A0%E9%80%9F%E5%8F%8A%E7%A7%BB%E5%8A%A8%E7%AB%AF%E9%83%A8%E7%BD%B2.md) 255 | 6. [Tips for building fast portrait segmentation network with TensorFlow Lite](https://hyperconnect.github.io/2018/07/06/tips-for-building-fast-portrait-segmentation-network-with-tensorflow-lite.html) 256 | 7. $\color{red}{*}$ [Literature On Neural network architecture ](https://www.automl.org/automl/literature-on-neural-architecture-search/) 257 | 8. [A 2019 Guide to Semantic Segmentation](https://heartbeat.fritz.ai/a-2019-guide-to-semantic-segmentation-ca8242f5a7fc) 258 | 9. [Semantic Segmentation](https://zhangbin0917.github.io/2018/09/18/Semantic-Segmentation/) 259 | 10. [Image segmentation in 2020: Architectures, Losses, Datasets, and Frameworks](https://neptune.ai/blog/image-segmentation-in-2020) 260 | 11. [New mobile neural network architectures](https://machinethink.net/blog/mobile-architectures/) 261 | 12. [NanoDet](https://zhuanlan.zhihu.com/p/306530300) 262 | 263 | 264 | 265 | ### Some projects 266 | 267 | 1. [Real-Time Video Segmentation on Mobile Devices](https://github.com/tantara/JejuNet) 268 | 2. [LightNet++: Boosted Light-weighted Networks for Real-time Semantic Segmentation](https://github.com/ansleliu/LightNetPlusPlus) 269 | 3. [mobile_phone_human_matting](https://github.com/lizhengwei1992/mobile_phone_human_matting) 270 | 4. [Real-Time Semantic Segmentation in Mobile device](https://github.com/akirasosa/mobile-semantic-segmentation) 271 | 5. [Semantic Segmentation on PyTorch](https://github.com/Tramac/awesome-semantic-segmentation-pytorch) 272 | 6. [Lightweight-Segmentation](https://github.com/Tramac/Lightweight-Segmentation) 273 | 7. [A PyTorch-Based Framework for Deep Learning in Computer Vision](https://github.com/donnyyou/torchcv) 274 | 8. [A high performance semantic segmentation toolkit based on PaddlePaddle](https://github.com/PaddlePaddle/PaddleSeg) 275 | 9. [PyTorch image models, scripts, pretrained weights](https://github.com/rwightman/pytorch-image-models) 276 | 10. [fast semantic segmentation models on CityScapes/Camvid DataSet by Pytorch](https://github.com/lxtGH/Fast_Seg) 277 | 11. [A semantic segmentation framework by pyotrch](https://github.com/Media-Smart/vedaseg) 278 | 12. [PyTorch for Semantic Segmentation](https://github.com/LikeLy-Journey/SegmenTron) 279 | 13. [murufeng/awesome_lightweight_networks](https://github.com/murufeng/awesome_lightweight_networks) 280 | 14. [sithu31296/semantic-segmentation](https://github.com/sithu31296/semantic-segmentation) 281 | 282 | 283 | 284 | 285 | ### Inference framework 286 | 287 | 1. ncnn 288 | 2. paddle 289 | 3. bolt 290 | 4. Mace 291 | 5. [Mnn]() 292 | 293 | --------------------------------------------------------------------------------