├── homographynet ├── __init__.py ├── losses.py ├── callbacks.py ├── data.py └── models.py ├── .gitattributes ├── model.png ├── demo ├── 000000084752.jpg ├── 000000111006.jpg ├── mobile-model-top.png ├── training-history-baseline.json ├── training-history-mobilenet.json └── demo_utils.py ├── models ├── homographynet_weights_tf_dim_ordering_tf_kernels.h5 └── mobile_homographynet_weights_tf_dim_ordering_tf_kernels.h5 ├── README.md ├── test.py ├── .gitignore ├── train.py ├── dataset ├── generate.py └── HOWTO.ipynb └── LICENSE /homographynet/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /.gitattributes: -------------------------------------------------------------------------------- 1 | models/*.h5 filter=lfs diff=lfs merge=lfs -text 2 | -------------------------------------------------------------------------------- /model.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/baudm/HomographyNet/HEAD/model.png -------------------------------------------------------------------------------- /demo/000000084752.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/baudm/HomographyNet/HEAD/demo/000000084752.jpg -------------------------------------------------------------------------------- /demo/000000111006.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/baudm/HomographyNet/HEAD/demo/000000111006.jpg -------------------------------------------------------------------------------- /demo/mobile-model-top.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/baudm/HomographyNet/HEAD/demo/mobile-model-top.png -------------------------------------------------------------------------------- /models/homographynet_weights_tf_dim_ordering_tf_kernels.h5: -------------------------------------------------------------------------------- 1 | version https://git-lfs.github.com/spec/v1 2 | oid sha256:915d92726132f3e1d38b69b64838ef2b5d8bbe8ea223b06c792aa72cce6030a6 3 | size 136851432 4 | -------------------------------------------------------------------------------- /models/mobile_homographynet_weights_tf_dim_ordering_tf_kernels.h5: -------------------------------------------------------------------------------- 1 | version https://git-lfs.github.com/spec/v1 2 | oid sha256:e161aabc5a04ff715a6f5706855a339d598d1216a4a5f45b90b8dbf5f8bcedc3 3 | size 13652544 4 | -------------------------------------------------------------------------------- /homographynet/losses.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | from keras import backend as K 4 | 5 | 6 | def mean_corner_error(y_true, y_pred): 7 | y_true = K.reshape(y_true, (-1, 4, 2)) 8 | y_pred = K.reshape(y_pred, (-1, 4, 2)) 9 | return K.mean(K.sqrt(K.sum(K.square(y_pred - y_true), axis=-1, keepdims=True)), axis=1) 10 | -------------------------------------------------------------------------------- /homographynet/callbacks.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | from keras.callbacks import Callback 4 | import keras.backend as K 5 | 6 | 7 | class LearningRateScheduler(Callback): 8 | """Learning rate scheduler. 9 | 10 | See Caffe SGD docs 11 | """ 12 | 13 | def __init__(self, base_lr, gamma, step_size): 14 | super().__init__() 15 | self._base_lr = base_lr 16 | self._gamma = gamma 17 | self._step_size = step_size 18 | self._steps = 0 19 | 20 | def on_epoch_begin(self, epoch, logs=None): 21 | self._steps = epoch * self.params['steps'] 22 | 23 | def on_batch_begin(self, batch, logs=None): 24 | self._steps += 1 25 | if self._steps % self._step_size == 0: 26 | exp = int(self._steps / self._step_size) 27 | lr = self._base_lr * (self._gamma ** exp) 28 | K.set_value(self.model.optimizer.lr, lr) 29 | print('New learning rate:', lr) 30 | -------------------------------------------------------------------------------- /demo/training-history-baseline.json: -------------------------------------------------------------------------------- 1 | {"val_loss": [0.19354102693498135, 0.12739295998179992, 0.10107517743015706, 0.07535269620550858, 0.07121625932991381, 0.06947115872351997, 0.06872109490941739, 0.06630724019831614, 0.06621885059401393, 0.06547791775796682, 0.06571078021079302, 0.06544167812950502], "loss": [0.2987717576057483, 0.18480461807205126, 0.14332890652693234, 0.11945557311463814, 0.10006262974192699, 0.09747491212322926, 0.09571483789632718, 0.0938057638781193, 0.09202016934465904, 0.09188166706989973, 0.09170804428175473, 0.09141863870219542], "mean_corner_error": [0.6954073146749765, 0.5365200017545467, 0.46612674883542915, 0.4218487001879093, 0.38343720784172036, 0.377853075980376, 0.37407092589598434, 0.3700219843784968, 0.36628736100135706, 0.3659157091188125, 0.365492082264943, 0.36477196804223916], "val_mean_corner_error": [0.5584322760502497, 0.4386297272230299, 0.38499146976688897, 0.32760320618664107, 0.31567529341567624, 0.31175987681549455, 0.30968311471066584, 0.303746140519751, 0.30370218594868975, 0.30143572725782036, 0.30219991609454155, 0.30164508293346465]} -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # HomographyNet 2 | Implementation of HomographyNet in Keras 3 | 4 | ## Project Organization 5 | ### dataset 6 | * Contains `generate.py`, a script for generating a suitable dataset for the HomographyNet model. 7 | * A Jupyter notebook with general info about the structure of the generated dataset. 8 | 9 | ### demo 10 | * Jupyter notebook with the training/test results 11 | * Raw image files used 12 | * Utility code used in the demo 13 | * Training history in JSON format 14 | 15 | ### homographynet 16 | A Python package containing the implementation of the HomographyNet model in Keras. 17 | 18 | ### models 19 | * `homographynet_weights_tf_dim_ordering_tf_kernels.h5` - pretrained weights of the baseline HomographyNet model 20 | * `mobile_homographynet_weights_tf_dim_ordering_tf_kernels.h5` - pretrained weights of the MobileNet-based model 21 | 22 | 23 | `test.py` - a script for evaluating a specified model, or the pretrained model. It will download and cache the weights of the pretrained model on first use. 24 | 25 | `train.py` - a script for training the HomographyNet model 26 | -------------------------------------------------------------------------------- /demo/training-history-mobilenet.json: -------------------------------------------------------------------------------- 1 | {"val_mean_corner_error": [0.35682278097699793, 0.21165128231048583, 0.1782944307592249, 0.1622980426682952, 0.11285762100170056, 0.12811730106025235, 0.11842101681673876, 0.09516570994009574, 0.09772256109301432, 0.1324321689536986, 0.08571089152246714, 0.0829087055051825], "val_loss": [0.09447163791278884, 0.03698192085139453, 0.0284453451369716, 0.026546657012444103, 0.010375709080447753, 0.015505577687208989, 0.012441847982236923, 0.007864549311343581, 0.00835219130921105, 0.018555171578348784, 0.006593521655692408, 0.005937991762394844], "loss": [0.2670581323180634, 0.039349728240870324, 0.019933056144688564, 0.013959162818530622, 0.011021304207806212, 0.009207057499398406, 0.008012221987801008, 0.007091656102028747, 0.006352777995634824, 0.005765775919963534, 0.005302959489815223, 0.004951145717014487], "mean_corner_error": [0.6185520167763416, 0.22406833454966546, 0.1580923172650047, 0.13195918211761193, 0.11717886615162476, 0.10733034630903067, 0.10017546369383733, 0.09448832359069433, 0.08977980569100533, 0.08571329606458163, 0.08240693285488165, 0.07967347254021427]} -------------------------------------------------------------------------------- /test.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | import os.path 4 | import sys 5 | 6 | from keras.models import load_model 7 | 8 | from homographynet import data 9 | from homographynet.models import create_model 10 | from homographynet.losses import mean_corner_error 11 | 12 | 13 | def main(): 14 | if len(sys.argv) > 2: 15 | name = os.path.basename(__file__) 16 | print('Usage: {} [trained model.h5]'.format(name)) 17 | exit(1) 18 | 19 | if len(sys.argv) == 2: 20 | model = load_model(sys.argv[1], compile=False) 21 | else: 22 | model = create_model(use_weights=True) 23 | 24 | model.summary() 25 | 26 | batch_size = 64 * 2 27 | 28 | loader = data.loader(data.TEST_PATH, batch_size) 29 | steps = int(data.TEST_SAMPLES / batch_size) 30 | 31 | # Optimizer doesn't matter in this case, we just want to set the loss and metrics 32 | model.compile('sgd', loss='mean_squared_error', metrics=[mean_corner_error]) 33 | evaluation = model.evaluate_generator(loader, steps) 34 | print('Test loss:', evaluation) 35 | 36 | 37 | if __name__ == '__main__': 38 | main() 39 | -------------------------------------------------------------------------------- /demo/demo_utils.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | import cv2 4 | import numpy as np 5 | 6 | from dataset.generate import * 7 | 8 | 9 | def mean_corner_error(y_true, y_pred): 10 | y_true = np.reshape(y_true, (-1, 4, 2)) 11 | y_pred = np.reshape(y_pred, (-1, 4, 2)) 12 | return np.mean(np.sqrt(np.sum(np.square(y_pred - y_true), axis=-1, keepdims=True)), axis=1) 13 | 14 | 15 | def draw_lines(img, points, color, thickness=2): 16 | out = img.copy() 17 | n = len(points) 18 | for i in range(n): 19 | j = (i + 1) % n 20 | cv2.line(out, tuple(points[i]), tuple(points[j]), color, thickness, cv2.LINE_AA) 21 | return out 22 | 23 | 24 | def transform_points(p_original, p_perturbed): 25 | # See: https://docs.opencv.org/2.4/modules/imgproc/doc/geometric_transformations.html#getperspectivetransform 26 | M = cv2.getPerspectiveTransform(np.float32(p_perturbed), np.float32(p_original)) 27 | # Convert to homogenous representation of points 28 | h_points = np.hstack((p_original, np.array([1, 1, 1, 1])[np.newaxis].T)) 29 | # Transform 30 | h_points = M.dot(h_points.T) 31 | t = h_points[2] 32 | h_points = h_points[:2] / t 33 | return h_points.T.astype('uint16') 34 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | env/ 12 | build/ 13 | develop-eggs/ 14 | dist/ 15 | downloads/ 16 | eggs/ 17 | .eggs/ 18 | lib/ 19 | lib64/ 20 | parts/ 21 | sdist/ 22 | var/ 23 | wheels/ 24 | *.egg-info/ 25 | .installed.cfg 26 | *.egg 27 | 28 | # PyInstaller 29 | # Usually these files are written by a python script from a template 30 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 31 | *.manifest 32 | *.spec 33 | 34 | # Installer logs 35 | pip-log.txt 36 | pip-delete-this-directory.txt 37 | 38 | # Unit test / coverage reports 39 | htmlcov/ 40 | .tox/ 41 | .coverage 42 | .coverage.* 43 | .cache 44 | nosetests.xml 45 | coverage.xml 46 | *.cover 47 | .hypothesis/ 48 | 49 | # Translations 50 | *.mo 51 | *.pot 52 | 53 | # Django stuff: 54 | *.log 55 | local_settings.py 56 | 57 | # Flask stuff: 58 | instance/ 59 | .webassets-cache 60 | 61 | # Scrapy stuff: 62 | .scrapy 63 | 64 | # Sphinx documentation 65 | docs/_build/ 66 | 67 | # PyBuilder 68 | target/ 69 | 70 | # Jupyter Notebook 71 | .ipynb_checkpoints 72 | 73 | # pyenv 74 | .python-version 75 | 76 | # celery beat schedule file 77 | celerybeat-schedule 78 | 79 | # SageMath parsed files 80 | *.sage.py 81 | 82 | # dotenv 83 | .env 84 | 85 | # virtualenv 86 | .venv 87 | venv/ 88 | ENV/ 89 | 90 | # Spyder project settings 91 | .spyderproject 92 | .spyproject 93 | 94 | # Rope project settings 95 | .ropeproject 96 | 97 | # mkdocs documentation 98 | /site 99 | 100 | # mypy 101 | .mypy_cache/ 102 | -------------------------------------------------------------------------------- /homographynet/data.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | import os.path 4 | import glob 5 | 6 | import numpy as np 7 | 8 | 9 | _SAMPLES_PER_ARCHIVE = 7680 10 | 11 | TRAIN_PATH = '/home/darwin/Projects/HomographyNet/repack' 12 | TRAIN_SAMPLES = 65 * _SAMPLES_PER_ARCHIVE 13 | 14 | TEST_PATH = '/home/darwin/Projects/HomographyNet/test-set' 15 | TEST_SAMPLES = 5 * _SAMPLES_PER_ARCHIVE 16 | 17 | 18 | def _shuffle_in_unison(a, b): 19 | """A hack to shuffle both a and b the same "random" way""" 20 | prng_state = np.random.get_state() 21 | np.random.shuffle(a) 22 | np.random.set_state(prng_state) 23 | np.random.shuffle(b) 24 | 25 | 26 | def loader(path, batch_size=64, normalize=True): 27 | """Generator to be used with model.fit_generator()""" 28 | while True: 29 | files = glob.glob(os.path.join(path, '*.npz')) 30 | np.random.shuffle(files) 31 | for npz in files: 32 | # Load pack into memory 33 | archive = np.load(npz) 34 | images = archive['images'] 35 | offsets = archive['offsets'] 36 | del archive 37 | _shuffle_in_unison(images, offsets) 38 | # Split into mini batches 39 | num_batches = int(len(offsets) / batch_size) 40 | images = np.array_split(images, num_batches) 41 | offsets = np.array_split(offsets, num_batches) 42 | while offsets: 43 | batch_images = images.pop() 44 | batch_offsets = offsets.pop() 45 | if normalize: 46 | batch_images = (batch_images - 127.5) / 127.5 47 | batch_offsets = batch_offsets / 32. 48 | yield batch_images, batch_offsets 49 | -------------------------------------------------------------------------------- /train.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | import os.path 4 | import sys 5 | 6 | import math 7 | 8 | from keras.callbacks import ModelCheckpoint 9 | from keras.models import load_model 10 | from keras.optimizers import SGD 11 | 12 | from homographynet import data 13 | from homographynet.callbacks import LearningRateScheduler 14 | from homographynet.losses import mean_corner_error 15 | from homographynet.models import create_model 16 | 17 | 18 | def main(): 19 | if len(sys.argv) > 2: 20 | name = os.path.basename(__file__) 21 | print('Usage: {} [existing model.h5]'.format(name)) 22 | exit(1) 23 | 24 | if len(sys.argv) == 2: 25 | model = load_model(sys.argv[1], compile=False) 26 | else: 27 | model = create_model() 28 | 29 | # Configuration 30 | batch_size = 64 31 | target_iterations = 90000 # at batch_size = 64 32 | base_lr = 0.005 33 | 34 | sgd = SGD(lr=base_lr, momentum=0.9) 35 | 36 | model.compile(optimizer=sgd, loss='mean_squared_error', metrics=[mean_corner_error]) 37 | model.summary() 38 | 39 | save_path = os.path.dirname(os.path.realpath(__file__)) 40 | checkpoint = ModelCheckpoint(os.path.join(save_path, 'model.{epoch:02d}.h5')) 41 | 42 | # LR scaling as described in the paper 43 | lr_scheduler = LearningRateScheduler(base_lr, 0.1, 30000) 44 | 45 | # In the paper, the 90,000 iterations was for batch_size = 64 46 | # So scale appropriately 47 | target_iterations = int(target_iterations * 64 / batch_size) 48 | # As stated in Keras docs 49 | steps_per_epoch = int(data.TRAIN_SAMPLES / batch_size) 50 | epochs = int(math.ceil(target_iterations / steps_per_epoch)) 51 | 52 | loader = data.loader(data.TRAIN_PATH, batch_size) 53 | 54 | val_loader = data.loader(data.TEST_PATH, batch_size) 55 | val_steps = int(data.TEST_SAMPLES / batch_size) 56 | 57 | # Train 58 | model.fit_generator(loader, steps_per_epoch, epochs, 59 | callbacks=[lr_scheduler, checkpoint], 60 | validation_data=val_loader, validation_steps=val_steps) 61 | 62 | 63 | if __name__ == '__main__': 64 | main() 65 | -------------------------------------------------------------------------------- /homographynet/models.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | import os.path 4 | 5 | from keras.applications import MobileNet 6 | from keras.utils.data_utils import get_file 7 | from keras.models import Sequential, Model 8 | from keras.layers import Conv2D, Dense, MaxPooling2D, InputLayer, Dropout, \ 9 | BatchNormalization, Flatten, Concatenate 10 | 11 | 12 | BASELINE_WEIGHTS_PATH = 'https://github.com/baudm/HomographyNet/raw/master/models/homographynet_weights_tf_dim_ordering_tf_kernels.h5' 13 | MOBILENET_WEIGHTS_PATH = 'https://github.com/baudm/HomographyNet/raw/master/models/mobile_homographynet_weights_tf_dim_ordering_tf_kernels.h5' 14 | 15 | 16 | def create_model(use_weights=False): 17 | model = Sequential(name='homographynet') 18 | model.add(InputLayer((128, 128, 2), name='input_1')) 19 | 20 | # 4 Layers with 64 filters, then another 4 with 128 filters 21 | filters = 4 * [64] + 4 * [128] 22 | for i, f in enumerate(filters, 1): 23 | model.add(Conv2D(f, 3, padding='same', activation='relu', name='conv2d_{}'.format(i))) 24 | model.add(BatchNormalization(name='batch_normalization_{}'.format(i))) 25 | # MaxPooling after every 2 Conv layers except the last one 26 | if i % 2 == 0 and i != 8: 27 | model.add(MaxPooling2D(strides=(2, 2), name='max_pooling2d_{}'.format(int(i/2)))) 28 | 29 | model.add(Flatten(name='flatten_1')) 30 | model.add(Dropout(0.5, name='dropout_1')) 31 | model.add(Dense(1024, activation='relu', name='dense_1')) 32 | model.add(Dropout(0.5, name='dropout_2')) 33 | 34 | # Regression model 35 | model.add(Dense(8, name='dense_2')) 36 | 37 | if use_weights: 38 | weights_name = os.path.basename(BASELINE_WEIGHTS_PATH) 39 | weights_path = get_file(weights_name, BASELINE_WEIGHTS_PATH, 40 | cache_subdir='models', 41 | file_hash='915d92726132f3e1d38b69b64838ef2b5d8bbe8ea223b06c792aa72cce6030a6') 42 | model.load_weights(weights_path) 43 | 44 | return model 45 | 46 | 47 | def create_mobilenet_model(use_weights=False): 48 | base_model = MobileNet(input_shape=(128, 128, 2), include_top=False, weights=None) 49 | # The output shape just before the pooling and dense layers is: (4, 4, 1024) 50 | x = base_model.output 51 | 52 | # 4 Conv layers in parallel with 2 4x4 filters each 53 | x = [Conv2D(2, 4, name='conv2d_{}'.format(i))(x) for i in range(1, 5)] 54 | x = Concatenate(name='concatenate_1')(x) 55 | x = Flatten(name='flatten_1')(x) 56 | 57 | model = Model(base_model.input, x, name='mobile_homographynet') 58 | 59 | if use_weights: 60 | weights_name = os.path.basename(MOBILENET_WEIGHTS_PATH) 61 | weights_path = get_file(weights_name, MOBILENET_WEIGHTS_PATH, 62 | cache_subdir='models', 63 | file_hash='e161aabc5a04ff715a6f5706855a339d598d1216a4a5f45b90b8dbf5f8bcedc3') 64 | model.load_weights(weights_path) 65 | 66 | return model 67 | -------------------------------------------------------------------------------- /dataset/generate.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | import sys 4 | import random 5 | import glob 6 | import os.path 7 | import uuid 8 | 9 | from queue import Queue, Empty 10 | from threading import Thread 11 | 12 | import numpy as np 13 | import cv2 14 | 15 | 16 | def scale_down(img, target_size): 17 | src_height, src_width = img.shape 18 | src_ratio = src_height/src_width 19 | target_width, target_height = target_size 20 | if src_ratio < target_height/target_width: 21 | dst_size = (int(np.round(target_height/src_ratio)), target_height) 22 | else: 23 | dst_size = (target_width, int(np.round(target_width*src_ratio))) 24 | return cv2.resize(img, dst_size, interpolation=cv2.INTER_AREA) 25 | 26 | 27 | def crop(img, origin, size): 28 | width, height = size 29 | x, y = origin 30 | return img[y:y + height, x:x + width] 31 | 32 | 33 | def center_crop(img, target_size): 34 | target_width, target_height = target_size 35 | # Note the reverse order of width and height 36 | height, width = img.shape 37 | x = int(np.round((width - target_width)/2)) 38 | y = int(np.round((height - target_height)/2)) 39 | return crop(img, (x, y), target_size) 40 | 41 | 42 | def generate_points(): 43 | # Choose top-left corner of patch (assume 0,0 is top-left of image) 44 | # Restrict points to within 24-px from the border 45 | p = 32 46 | x, y = (random.randint(56, 136), 56) 47 | patch = [ 48 | (x, y), 49 | (x + 128, y), 50 | (x + 128, y + 128), 51 | (x, y + 128) 52 | ] 53 | # Perturb 54 | perturbed_patch = [(x + random.randint(-p, p), y + random.randint(-p, p)) for x, y in patch] 55 | return np.array(patch), np.array(perturbed_patch) 56 | 57 | 58 | def warp(img, orig, perturbed, target_size): 59 | # Get inverse homography matrix 60 | M = cv2.getPerspectiveTransform(np.float32(perturbed), np.float32(orig)) 61 | return cv2.warpPerspective(img, M, target_size, flags=cv2.INTER_CUBIC) 62 | 63 | 64 | def process_image(image_path, num_output=1): 65 | # Read as grayscale 66 | img = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE) 67 | if img.shape < (240, 320): 68 | return 69 | 70 | target_size = (320, 240) 71 | img = scale_down(img, target_size) 72 | img = center_crop(img, target_size) 73 | 74 | patch_size = (128, 128) 75 | image_pairs = [] 76 | offsets = [] 77 | #orig_points = [] 78 | #perturbed_points = [] 79 | while len(offsets) < num_output: 80 | orig, perturbed = generate_points() 81 | a = crop(img, orig[0], patch_size) 82 | b = warp(img, orig, perturbed, target_size) 83 | b = crop(b, orig[0], patch_size) 84 | try: 85 | d = np.stack((a, b), axis=-1) 86 | except ValueError: 87 | continue 88 | image_pairs.append(d) 89 | offset = (perturbed - orig).reshape(-1) 90 | offsets.append(offset) 91 | #orig_points.append(orig) 92 | #perturbed_points.append(perturbed) 93 | print('done:', image_path) 94 | return image_pairs, offsets 95 | 96 | 97 | class Worker(Thread): 98 | 99 | def __init__(self, input_queue, output_queue, num_samples): 100 | Thread.__init__(self) 101 | self.input_queue = input_queue 102 | self.output_queue = output_queue 103 | self.num_samples = num_samples 104 | 105 | def run(self): 106 | while True: 107 | img_path = self.input_queue.get() 108 | if img_path is None: 109 | break 110 | output = process_image(img_path, self.num_samples) 111 | self.input_queue.task_done() 112 | if output is not None: 113 | self.output_queue.put(output) 114 | 115 | 116 | def pack(outdir, image_pairs, offsets): 117 | name = str(uuid.uuid4()) 118 | pack = os.path.join(outdir, name + '.npz') 119 | with open(pack, 'wb') as f: 120 | np.savez(f, images=np.stack(image_pairs), offsets=np.stack(offsets)) 121 | print('bundled:', name) 122 | 123 | 124 | def bundle(queue, outdir): 125 | image_pairs = [] 126 | offsets = [] 127 | #orig_points = [] 128 | #perturbed_points = [] 129 | while True: 130 | try: 131 | d, o = queue.get(timeout=10) 132 | except Empty: 133 | break 134 | image_pairs.extend(d) 135 | offsets.extend(o) 136 | #orig_points.extend(orig) 137 | #perturbed_points.extend(perturbed) 138 | 139 | if len(image_pairs) >= 7680: 140 | pack(outdir, image_pairs, offsets) 141 | image_pairs = [] 142 | offsets = [] 143 | queue.task_done() 144 | 145 | if image_pairs: 146 | pack(outdir, image_pairs, offsets) 147 | 148 | 149 | def main(): 150 | if len(sys.argv) < 4: 151 | print('Usage: generate.py ') 152 | exit(1) 153 | output_dir = sys.argv[1] 154 | samples = int(sys.argv[2]) 155 | input_dirs = sys.argv[3:] 156 | 157 | # Create a queue to communicate with the worker threads 158 | input_queue = Queue() 159 | output_queue = Queue() 160 | 161 | num_workers = 8 162 | workers = [] 163 | # Create worker threads 164 | for i in range(num_workers): 165 | worker = Worker(input_queue, output_queue, samples) 166 | worker.start() 167 | workers.append(worker) 168 | 169 | for d in input_dirs: 170 | for i in glob.iglob(os.path.join(d, '*.jpg')): 171 | input_queue.put(i) 172 | 173 | bundle(output_queue, output_dir) 174 | 175 | input_queue.join() 176 | for i in range(num_workers): 177 | input_queue.put(None) 178 | for worker in workers: 179 | worker.join() 180 | 181 | 182 | if __name__ == '__main__': 183 | main() 184 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "{}" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright {yyyy} {name of copyright owner} 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /dataset/HOWTO.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "markdown", 5 | "metadata": {}, 6 | "source": [ 7 | "# About\n", 8 | "This data set was generated using images from the [COCO 2017 Test Images (~41k images)](http://images.cocodataset.org/zips/test2017.zip) using the method described in [HomographyNet](https://arxiv.org/pdf/1606.03798.pdf).\n", 9 | "\n", 10 | "## Sample\n", 11 | "A sample consists of a pair of 128x128 patches and their corresponding corner offsets (4 pairs). Note that samples are not normalized in order to save disk space (33kB for 1 unnormalized sample vs 233kB)\n", 12 | "\n", 13 | "More specifically, a sample is a tuple of a (128, 128, 2) NumPy array (128x128 grayscale images stacked channel-wise) and a (8,) NumPy array (8-element vector containing the perturbations applied to the 4 corners).\n", 14 | "\n", 15 | "Each image is sampled 12 times on average generating a total of 485,376 samples in total.\n", 16 | "\n", 17 | "## Bundles\n", 18 | "For efficient loading, samples are bundled together in NumPy archives (.npz) but are uncompressed because the compression ratio is too low to make the additional CPU time worth it. Packing is done sequentially. That is, samples from a single image are grouped together. An archive is about 96 MiB in size and contains 3,072 samples.\n", 19 | "\n", 20 | "In this machine, a single archive can be read from disk (HDD) and loaded into memory in about 1.33s (first load, uncached). On succeeding loads (already cached), loading into memory takes about 200ms. This is much faster than having to load images individually on the fly (the HDD/CPU will bottleneck the GPU). With a batch size of 64, each archive is good for 48 iterations before having to load the next one (decreasing overall disk access and CPU wait time).\n", 21 | "\n", 22 | "## Downloading\n", 23 | "The files are hosted [here](http://192.168.1.126/~darwin/coco2017/). An archive's filename is also its MD5 hash, so you can easily check for file integrity.\n", 24 | "\n", 25 | "You can easily download all the archives using wget:\n", 26 | "\n", 27 | "`wget -rcnd -l 1 http://192.168.1.126/~darwin/coco2017/`\n", 28 | "\n", 29 | "There are 158 archives in all, or about 15GiB of data." 30 | ] 31 | }, 32 | { 33 | "cell_type": "code", 34 | "execution_count": 141, 35 | "metadata": { 36 | "collapsed": false 37 | }, 38 | "outputs": [ 39 | { 40 | "name": "stdout", 41 | "output_type": "stream", 42 | "text": [ 43 | "keys: ['images', 'offsets']\n", 44 | "images.shape: (3072, 128, 128, 2)\n", 45 | "offsets.shape: (3072, 8)\n", 46 | "sample images shape: (128, 128, 2)\n", 47 | "(64, 128, 128, 2)\n" 48 | ] 49 | } 50 | ], 51 | "source": [ 52 | "%matplotlib inline\n", 53 | "\n", 54 | "import numpy as np\n", 55 | "import matplotlib.pyplot as plt\n", 56 | "\n", 57 | "path = '/home/darwin/Projects/datasets/packed/f01906891b1f8ea2e2715d9f99f126f6.npz'\n", 58 | "\n", 59 | "archive = np.load(path)\n", 60 | "print('keys:', archive.files)\n", 61 | "\n", 62 | "images = archive['images']\n", 63 | "offsets = archive['offsets']\n", 64 | "\n", 65 | "print('images.shape:', images.shape)\n", 66 | "print('offsets.shape:', offsets.shape)\n", 67 | "\n", 68 | "sample_idx = 2335\n", 69 | "\n", 70 | "print('sample images shape:', images[sample_idx].shape)\n", 71 | "\n", 72 | "a=np.split(images,48)\n", 73 | "print(a[0].shape)" 74 | ] 75 | }, 76 | { 77 | "cell_type": "code", 78 | "execution_count": 136, 79 | "metadata": { 80 | "collapsed": false 81 | }, 82 | "outputs": [ 83 | { 84 | "data": { 85 | "text/plain": [ 86 | "" 87 | ] 88 | }, 89 | "execution_count": 136, 90 | "metadata": {}, 91 | "output_type": "execute_result" 92 | }, 93 | { 94 | "data": { 95 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQUAAAD8CAYAAAB+fLH0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUKBAgRSKRaFAgQIpFItCgQIFUvj/zMja8SGv5ucAAAAASUVORK5CYII=\n", 96 | "text/plain": [ 97 | "" 98 | ] 99 | }, 100 | "metadata": {}, 101 | "output_type": "display_data" 102 | } 103 | ], 104 | "source": [ 105 | "img = np.split(images[sample_idx], 2, axis=-1)\n", 106 | "plt.imshow(img[0].squeeze(), cmap='gray')" 107 | ] 108 | }, 109 | { 110 | "cell_type": "code", 111 | "execution_count": 137, 112 | "metadata": { 113 | "collapsed": false 114 | }, 115 | "outputs": [ 116 | { 117 | "data": { 118 | "text/plain": [ 119 | "" 120 | ] 121 | }, 122 | "execution_count": 137, 123 | "metadata": {}, 124 | "output_type": "execute_result" 125 | }, 126 | { 127 | "data": { 128 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQUAAAD8CAYAAAB+fLH0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n6g0ZAAAAAElFTkSuQmCC\n", 129 | "text/plain": [ 130 | "" 131 | ] 132 | }, 133 | "metadata": {}, 134 | "output_type": "display_data" 135 | } 136 | ], 137 | "source": [ 138 | "plt.imshow(img[1].squeeze(), cmap='gray')" 139 | ] 140 | }, 141 | { 142 | "cell_type": "code", 143 | "execution_count": 138, 144 | "metadata": { 145 | "collapsed": false 146 | }, 147 | "outputs": [ 148 | { 149 | "name": "stdout", 150 | "output_type": "stream", 151 | "text": [ 152 | "[ 24 4 26 16 -9 -17 20 20]\n" 153 | ] 154 | } 155 | ], 156 | "source": [ 157 | "# The order (although it doesn't matter) is: top-left, bottom-left, bottom-right, top-right\n", 158 | "print(offsets[sample_idx])" 159 | ] 160 | }, 161 | { 162 | "cell_type": "code", 163 | "execution_count": 34, 164 | "metadata": { 165 | "collapsed": false 166 | }, 167 | "outputs": [], 168 | "source": [ 169 | "# Efficient loading in Keras using a Python generator\n", 170 | "\n", 171 | "import os.path\n", 172 | "import glob\n", 173 | "\n", 174 | "import numpy as np\n", 175 | "\n", 176 | "def data_loader(path, batch_size=64):\n", 177 | " \"\"\"Generator to be used with model.fit_generator()\"\"\"\n", 178 | " while True:\n", 179 | " for npz in glob.glob(os.path.join(path, '*.npz')):\n", 180 | " # Load pack into memory\n", 181 | " archive = np.load(npz)\n", 182 | " images = archive['images']\n", 183 | " offsets = archive['offsets']\n", 184 | " # Yield minibatch\n", 185 | " for i in range(0, len(offsets), batch_size):\n", 186 | " end_i = i + batch_size\n", 187 | " try:\n", 188 | " batch_images = images[i:end_i]\n", 189 | " batch_offsets = offsets[i:end_i]\n", 190 | " except IndexError:\n", 191 | " continue\n", 192 | " # Normalize\n", 193 | " batch_images = (batch_images - 127.5) / 127.5\n", 194 | " batch_offsets = batch_offsets / 32.\n", 195 | " yield batch_images, batch_offsets\n", 196 | "\n", 197 | "# Dataset-specific\n", 198 | "train_data_path = '/path/to/training-data'\n", 199 | "test_data_path = '/path/to/test-data'\n", 200 | "num_samples = 150 * 3072 # 158 archives x 3,072 samples per archive, but use just 150 and save the 8 for testing\n", 201 | "\n", 202 | "# From the paper\n", 203 | "batch_size = 64\n", 204 | "total_iterations = 90000\n", 205 | "\n", 206 | "steps_per_epoch = num_samples / batch_size # As stated in Keras docs\n", 207 | "epochs = int(total_iterations / steps_per_epoch)\n", 208 | "\n", 209 | "# model is some Keras Model instance\n", 210 | "\n", 211 | "# Train\n", 212 | "model.fit_generator(data_loader(train_data_path, batch_size),\n", 213 | " steps_per_epoch=steps_per_epoch,\n", 214 | " epochs=epochs)\n", 215 | "\n", 216 | "# Test\n", 217 | "model.evaluate_generator(data_loader(test_data_path, 1),\n", 218 | " steps=100)" 219 | ] 220 | }, 221 | { 222 | "cell_type": "code", 223 | "execution_count": null, 224 | "metadata": { 225 | "collapsed": true 226 | }, 227 | "outputs": [], 228 | "source": [] 229 | } 230 | ], 231 | "metadata": { 232 | "kernelspec": { 233 | "display_name": "Python 3", 234 | "language": "python", 235 | "name": "python3" 236 | }, 237 | "language_info": { 238 | "codemirror_mode": { 239 | "name": "ipython", 240 | "version": 3 241 | }, 242 | "file_extension": ".py", 243 | "mimetype": "text/x-python", 244 | "name": "python", 245 | "nbconvert_exporter": "python", 246 | "pygments_lexer": "ipython3", 247 | "version": "3.5.3" 248 | } 249 | }, 250 | "nbformat": 4, 251 | "nbformat_minor": 0 252 | } 253 | --------------------------------------------------------------------------------