def load_all(config, args, mode):
features_path = get_results_path(config.results_path, 'features',
args.prefix, mode)
image_patches = np.load(features_path / 'image_patches.npy')
feature_matrix = np.load(features_path / 'feature_matrix.npy')
labels = np.load(features_path / 'labels.npy')
return image_patches, feature_matrix, labels
if __name__ == '__main__':
logger = get_logger('visual_histograms')
set_excepthook(logger)
args = parse_arguments()
config = load_config(args.config)
set_seed(config.seed)
model_type = 'bow' if args.bow else 'fv'
CLUSTERS_NUM = 10
# train_image_patches, train_feature_matrix, train_labels = load_all(config, args, 'train')
test_image_patches, test_feature_matrix, test_labels = load_all(
config, args, 'test')
model_path = get_results_path(config.results_path, model_type, args.prefix,
str(CLUSTERS_NUM))
model = joblib.load(model_path / 'best_model.pkl').best_estimator_
transformer_name = 'bag_of_words' if args.bow else 'fisher_vector'
transformer = model.named_steps[transformer_name]
# train_points = transformer.transform(train_feature_matrix)
test_points = transformer.transform(test_feature_matrix)
# compute distances from train and test to gmm clusters
# train_distances = cdist(
# train_feature_matrix.reshape(-1, 256), transformer.gmm_[0].transpose())
# test_distances = cdist(
# test_feature_matrix.reshape(-1, 256), transformer.gmm_[0].transpose())
# plot_similarity_mosaic(test_distances, test_image_patches, model_path)
stats = dict(zip(unique.tolist(), counts.tolist()))
with open(path, mode='w') as f:
yaml.dump(stats, f)
if __name__ == '__main__':
logger = get_logger('extract_features')
set_excepthook(logger)
args = parse_arguments()
config = load_config(args.config)
set_seed(config.seed)
mode = 'test' if args.test else 'train'
if args.augment:
args.prefix += '_aug'
results_path = get_results_path(config.results_path, args.model,
args.prefix, mode)
logger.info(('Extracting features...\n'
'prefix: %s\n'
'mode: %s\n'
'augmentation: %s\n'
'model: %s'), args.prefix, mode, args.augment, args.model)
transform = [
NumpyToTensor(),
get_normalization_transform(),
]
if args.augment:
transform.insert(0, NumpyGaussianNoise(sigma=0.01))
transform = Compose(transform)
augmentation = None
help='path to python module with shared experiment configuration')
parser.add_argument('--augment', action='store_true',
help='enable augmentation')
parser.add_argument('--features', default='alexnet',
help='which feature extraction method to use; can be one of alexnet, resnet18, inceptionv3')
return parser.parse_args()
if __name__ == '__main__':
logger = get_logger('hyperparameters')
set_excepthook(logger)
args = parse_arguments()
config = load_config(args.config)
set_seed(config.seed)
features_path = get_results_path(
config.results_path, args.features, args.prefix, 'train')
if args.augment:
args.prefix += '_aug'
aug_features_path = get_results_path(
config.results_path, args.features, args.prefix, 'train')
train_results_path = get_results_path(
config.results_path, args.features, str(args.model) + '_' + str(args.prefix), 'train')
train_results_path.mkdir(parents=True, exist_ok=True)
logger.info('Fitting hyperparameters for prefix %s with %s model',
args.prefix, args.model)
feature_matrix = np.load(features_path / 'feature_matrix.npy')
labels = np.load(features_path / 'labels.npy')
if args.augment:
aug_feature_matrix = np.load(aug_features_path / 'feature_matrix.npy')
aug_labels = np.load(aug_features_path / 'labels.npy')
default=50,
type=int,
help='clusters number to train the model')
return parser.parse_args()
if __name__ == '__main__':
logger = get_logger('train_model')
set_excepthook(logger)
args = parse_arguments()
config = load_config(args.config)
set_seed(config.seed)
model = 'bow' if args.bow else 'fv'
train_features_path = get_results_path(config.results_path, 'features',
args.prefix, 'train')
logger.info('Fitting model...') # add config
feature_matrix = np.load(train_features_path / 'feature_matrix.npy')
labels = np.load(train_features_path / 'labels.npy')
pipeline = bow if args.bow else fisher_vector_transformer
if args.bow:
param_grid = config.bow_param_grid
for i in range(2):
param_grid[i]['bag_of_words__clusters_number'] = [args.clusters]
else:
param_grid = config.fv_param_grid
for i in range(2):
param_grid[i]['fisher_vector__gmm_clusters_number'] = [
args.clusters
]
pipeline = model_selection.GridSearchCV(pipeline, param_grid, n_jobs=24)