def get_training_transforms(self): assert self.params.get( 'mirror' ) is None, "old version of params, use new keyword do_mirror" tr_transforms = [] if self.params.get("selected_data_channels"): tr_transforms.append( DataChannelSelectionTransform( self.params.get("selected_data_channels"))) if self.params.get("selected_seg_channels"): tr_transforms.append( SegChannelSelectionTransform( self.params.get("selected_seg_channels"))) # don't do color augmentations while in 2d mode with 3d data because the color channel is overloaded!! if self.params.get("dummy_2D", False): ignore_axes = (0, ) tr_transforms.append(Convert3DTo2DTransform()) else: ignore_axes = None tr_transforms.append( SpatialTransform( self._spatial_transform_patch_size, patch_center_dist_from_border=None, do_elastic_deform=self.params.get("do_elastic"), alpha=self.params.get("elastic_deform_alpha"), sigma=self.params.get("elastic_deform_sigma"), do_rotation=self.params.get("do_rotation"), angle_x=self.params.get("rotation_x"), angle_y=self.params.get("rotation_y"), angle_z=self.params.get("rotation_z"), do_scale=self.params.get("do_scaling"), scale=self.params.get("scale_range"), order_data=self.params.get("order_data"), border_mode_data=self.params.get("border_mode_data"), border_cval_data=self.params.get("border_cval_data"), order_seg=self.params.get("order_seg"), border_mode_seg=self.params.get("border_mode_seg"), border_cval_seg=self.params.get("border_cval_seg"), random_crop=self.params.get("random_crop"), p_el_per_sample=self.params.get("p_eldef"), p_scale_per_sample=self.params.get("p_scale"), p_rot_per_sample=self.params.get("p_rot"), independent_scale_for_each_axis=self.params.get( "independent_scale_factor_for_each_axis"), )) if self.params.get("dummy_2D"): tr_transforms.append(Convert2DTo3DTransform()) # we need to put the color augmentations after the dummy 2d part (if applicable). Otherwise the overloaded color # channel gets in the way tr_transforms.append(GaussianNoiseTransform(p_per_sample=0.15)) tr_transforms.append( GaussianBlurTransform((0.5, 1.5), different_sigma_per_channel=True, p_per_sample=0.2, p_per_channel=0.5), ) tr_transforms.append( BrightnessMultiplicativeTransform(multiplier_range=(0.75, 1.3), p_per_sample=0.15)) if self.params.get("do_additive_brightness"): tr_transforms.append( BrightnessTransform( self.params.get("additive_brightness_mu"), self.params.get("additive_brightness_sigma"), True, p_per_sample=self.params.get( "additive_brightness_p_per_sample"), p_per_channel=self.params.get( "additive_brightness_p_per_channel"))) tr_transforms.append( ContrastAugmentationTransform(contrast_range=(0.65, 1.5), p_per_sample=0.15)) tr_transforms.append( SimulateLowResolutionTransform(zoom_range=(0.5, 1), per_channel=True, p_per_channel=0.5, order_downsample=0, order_upsample=3, p_per_sample=0.25, ignore_axes=ignore_axes), ) tr_transforms.append( GammaTransform(self.params.get("gamma_range"), True, True, retain_stats=self.params.get("gamma_retain_stats"), p_per_sample=0.15)) # inverted gamma if self.params.get("do_gamma"): tr_transforms.append( GammaTransform( self.params.get("gamma_range"), False, True, retain_stats=self.params.get("gamma_retain_stats"), p_per_sample=self.params["p_gamma"])) if self.params.get("do_mirror") or self.params.get("mirror"): tr_transforms.append( MirrorTransform(self.params.get("mirror_axes"))) if self.params.get("use_mask_for_norm"): use_mask_for_norm = self.params.get("use_mask_for_norm") tr_transforms.append( MaskTransform(use_mask_for_norm, mask_idx_in_seg=0, set_outside_to=0)) tr_transforms.append(RemoveLabelTransform(-1, 0)) tr_transforms.append(RenameTransform('seg', 'target', True)) tr_transforms.append(NumpyToTensor(['data', 'target'], 'float')) return Compose(tr_transforms)
def get_train_transforms(self) -> List[AbstractTransform]: # used for transpost and rot90 matching_axes = np.array( [sum([i == j for j in self.patch_size]) for i in self.patch_size]) valid_axes = list(np.where(matching_axes == np.max(matching_axes))[0]) tr_transforms = [] if self.data_aug_params['selected_seg_channels'] is not None: tr_transforms.append( SegChannelSelectionTransform( self.data_aug_params['selected_seg_channels'])) if self.do_dummy_2D_aug: ignore_axes = (0, ) tr_transforms.append(Convert3DTo2DTransform()) patch_size_spatial = self.patch_size[1:] else: patch_size_spatial = self.patch_size ignore_axes = None tr_transforms.append( SpatialTransform( patch_size_spatial, patch_center_dist_from_border=None, do_elastic_deform=False, do_rotation=True, angle_x=self.data_aug_params["rotation_x"], angle_y=self.data_aug_params["rotation_y"], angle_z=self.data_aug_params["rotation_z"], p_rot_per_axis=0.5, do_scale=True, scale=self.data_aug_params['scale_range'], border_mode_data="constant", border_cval_data=0, order_data=3, border_mode_seg="constant", border_cval_seg=-1, order_seg=1, random_crop=False, p_el_per_sample=0.2, p_scale_per_sample=0.2, p_rot_per_sample=0.4, independent_scale_for_each_axis=True, )) if self.do_dummy_2D_aug: tr_transforms.append(Convert2DTo3DTransform()) if np.any(matching_axes > 1): tr_transforms.append( Rot90Transform((0, 1, 2, 3), axes=valid_axes, data_key='data', label_key='seg', p_per_sample=0.5), ) if np.any(matching_axes > 1): tr_transforms.append( TransposeAxesTransform(valid_axes, data_key='data', label_key='seg', p_per_sample=0.5)) tr_transforms.append( OneOfTransform([ MedianFilterTransform((2, 8), same_for_each_channel=False, p_per_sample=0.2, p_per_channel=0.5), GaussianBlurTransform((0.3, 1.5), different_sigma_per_channel=True, p_per_sample=0.2, p_per_channel=0.5) ])) tr_transforms.append(GaussianNoiseTransform(p_per_sample=0.1)) tr_transforms.append( BrightnessTransform(0, 0.5, per_channel=True, p_per_sample=0.1, p_per_channel=0.5)) tr_transforms.append( OneOfTransform([ ContrastAugmentationTransform(contrast_range=(0.5, 2), preserve_range=True, per_channel=True, data_key='data', p_per_sample=0.2, p_per_channel=0.5), ContrastAugmentationTransform(contrast_range=(0.5, 2), preserve_range=False, per_channel=True, data_key='data', p_per_sample=0.2, p_per_channel=0.5), ])) tr_transforms.append( SimulateLowResolutionTransform(zoom_range=(0.25, 1), per_channel=True, p_per_channel=0.5, order_downsample=0, order_upsample=3, p_per_sample=0.15, ignore_axes=ignore_axes)) tr_transforms.append( GammaTransform((0.7, 1.5), invert_image=True, per_channel=True, retain_stats=True, p_per_sample=0.1)) tr_transforms.append( GammaTransform((0.7, 1.5), invert_image=True, per_channel=True, retain_stats=True, p_per_sample=0.1)) if self.do_mirroring: tr_transforms.append(MirrorTransform(self.mirror_axes)) tr_transforms.append( BlankRectangleTransform([[max(1, p // 10), p // 3] for p in self.patch_size], rectangle_value=np.mean, num_rectangles=(1, 5), force_square=False, p_per_sample=0.4, p_per_channel=0.5)) tr_transforms.append( BrightnessGradientAdditiveTransform( lambda x, y: np.exp( np.random.uniform(np.log(x[y] // 6), np.log(x[y]))), (-0.5, 1.5), max_strength=lambda x, y: np.random.uniform(-5, -1) if np.random.uniform() < 0.5 else np.random.uniform(1, 5), mean_centered=False, same_for_all_channels=False, p_per_sample=0.3, p_per_channel=0.5)) tr_transforms.append( LocalGammaTransform( lambda x, y: np.exp( np.random.uniform(np.log(x[y] // 6), np.log(x[y]))), (-0.5, 1.5), lambda: np.random.uniform(0.01, 0.8) if np.random.uniform() < 0.5 else np.random.uniform(1.5, 4), same_for_all_channels=False, p_per_sample=0.3, p_per_channel=0.5)) tr_transforms.append( SharpeningTransform(strength=(0.1, 1), same_for_each_channel=False, p_per_sample=0.2, p_per_channel=0.5)) if any(self.use_mask_for_norm.values()): tr_transforms.append( MaskTransform(self.use_mask_for_norm, mask_idx_in_seg=0, set_outside_to=0)) tr_transforms.append(RemoveLabelTransform(-1, 0)) if self.data_aug_params["move_last_seg_chanel_to_data"]: all_class_labels = np.arange(1, self.num_classes) tr_transforms.append( MoveSegAsOneHotToData(1, all_class_labels, 'seg', 'data')) if self.data_aug_params["cascade_do_cascade_augmentations"]: tr_transforms.append( ApplyRandomBinaryOperatorTransform(channel_idx=list( range(-len(all_class_labels), 0)), p_per_sample=0.4, key="data", strel_size=(1, 8), p_per_label=1)) tr_transforms.append( RemoveRandomConnectedComponentFromOneHotEncodingTransform( channel_idx=list(range(-len(all_class_labels), 0)), key="data", p_per_sample=0.2, fill_with_other_class_p=0.15, dont_do_if_covers_more_than_X_percent=0)) tr_transforms.append(RenameTransform('seg', 'target', True)) if self.regions is not None: tr_transforms.append( ConvertSegmentationToRegionsTransform(self.regions, 'target', 'target')) if self.deep_supervision_scales is not None: tr_transforms.append( DownsampleSegForDSTransform2(self.deep_supervision_scales, 0, input_key='target', output_key='target')) tr_transforms.append(NumpyToTensor(['data', 'target'], 'float')) return tr_transforms
def get_moreDA_augmentation(dataloader_train, dataloader_val, patch_size, params=default_3D_augmentation_params, border_val_seg=-1, seeds_train=None, seeds_val=None, order_seg=1, order_data=3, deep_supervision_scales=None, soft_ds=False, classes=None, pin_memory=True): assert params.get( 'mirror') is None, "old version of params, use new keyword do_mirror" tr_transforms = [] if params.get("selected_data_channels") is not None: tr_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) if params.get("selected_seg_channels") is not None: tr_transforms.append( SegChannelSelectionTransform(params.get("selected_seg_channels"))) # don't do color augmentations while in 2d mode with 3d data because the color channel is overloaded!! if params.get("dummy_2D") is not None and params.get("dummy_2D"): ignore_axes = (0, ) tr_transforms.append(Convert3DTo2DTransform()) else: ignore_axes = None tr_transforms.append( SpatialTransform(patch_size, patch_center_dist_from_border=None, do_elastic_deform=params.get("do_elastic"), alpha=params.get("elastic_deform_alpha"), sigma=params.get("elastic_deform_sigma"), do_rotation=params.get("do_rotation"), angle_x=params.get("rotation_x"), angle_y=params.get("rotation_y"), angle_z=params.get("rotation_z"), p_rot_per_axis=params.get("rotation_p_per_axis"), do_scale=params.get("do_scaling"), scale=params.get("scale_range"), border_mode_data=params.get("border_mode_data"), border_cval_data=0, order_data=order_data, border_mode_seg="constant", border_cval_seg=border_val_seg, order_seg=order_seg, random_crop=params.get("random_crop"), p_el_per_sample=params.get("p_eldef"), p_scale_per_sample=params.get("p_scale"), p_rot_per_sample=params.get("p_rot"), independent_scale_for_each_axis=params.get( "independent_scale_factor_for_each_axis"))) if params.get("dummy_2D"): tr_transforms.append(Convert2DTo3DTransform()) # we need to put the color augmentations after the dummy 2d part (if applicable). Otherwise the overloaded color # channel gets in the way tr_transforms.append(GaussianNoiseTransform(p_per_sample=0.1)) tr_transforms.append( GaussianBlurTransform((0.5, 1.), different_sigma_per_channel=True, p_per_sample=0.2, p_per_channel=0.5)) tr_transforms.append( BrightnessMultiplicativeTransform(multiplier_range=(0.75, 1.25), p_per_sample=0.15)) if params.get("do_additive_brightness"): tr_transforms.append( BrightnessTransform( params.get("additive_brightness_mu"), params.get("additive_brightness_sigma"), True, p_per_sample=params.get("additive_brightness_p_per_sample"), p_per_channel=params.get("additive_brightness_p_per_channel"))) tr_transforms.append(ContrastAugmentationTransform(p_per_sample=0.15)) tr_transforms.append( SimulateLowResolutionTransform(zoom_range=(0.5, 1), per_channel=True, p_per_channel=0.5, order_downsample=0, order_upsample=3, p_per_sample=0.25, ignore_axes=ignore_axes)) tr_transforms.append( GammaTransform(params.get("gamma_range"), True, True, retain_stats=params.get("gamma_retain_stats"), p_per_sample=0.1)) # inverted gamma if params.get("do_gamma"): tr_transforms.append( GammaTransform(params.get("gamma_range"), False, True, retain_stats=params.get("gamma_retain_stats"), p_per_sample=params["p_gamma"])) if params.get("do_mirror") or params.get("mirror"): tr_transforms.append(MirrorTransform(params.get("mirror_axes"))) if params.get("mask_was_used_for_normalization") is not None: mask_was_used_for_normalization = params.get( "mask_was_used_for_normalization") tr_transforms.append( MaskTransform(mask_was_used_for_normalization, mask_idx_in_seg=0, set_outside_to=0)) tr_transforms.append(RemoveLabelTransform(-1, 0)) if params.get("move_last_seg_chanel_to_data") is not None and params.get( "move_last_seg_chanel_to_data"): tr_transforms.append( MoveSegAsOneHotToData(1, params.get("all_segmentation_labels"), 'seg', 'data')) if params.get( "cascade_do_cascade_augmentations") is not None and params.get( "cascade_do_cascade_augmentations"): if params.get("cascade_random_binary_transform_p") > 0: tr_transforms.append( ApplyRandomBinaryOperatorTransform( channel_idx=list( range(-len(params.get("all_segmentation_labels")), 0)), p_per_sample=params.get( "cascade_random_binary_transform_p"), key="data", strel_size=params.get( "cascade_random_binary_transform_size"), p_per_label=params.get( "cascade_random_binary_transform_p_per_label"))) if params.get("cascade_remove_conn_comp_p") > 0: tr_transforms.append( RemoveRandomConnectedComponentFromOneHotEncodingTransform( channel_idx=list( range(-len(params.get("all_segmentation_labels")), 0)), key="data", p_per_sample=params.get("cascade_remove_conn_comp_p"), fill_with_other_class_p=params.get( "cascade_remove_conn_comp_max_size_percent_threshold" ), dont_do_if_covers_more_than_X_percent=params.get( "cascade_remove_conn_comp_fill_with_other_class_p") )) tr_transforms.append(RenameTransform('seg', 'target', True)) if deep_supervision_scales is not None: if soft_ds: assert classes is not None tr_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'target', 'target', classes)) else: tr_transforms.append( DownsampleSegForDSTransform2(deep_supervision_scales, 0, 0, input_key='target', output_key='target')) tr_transforms.append(NumpyToTensor(['data', 'target'], 'float')) tr_transforms = Compose(tr_transforms) batchgenerator_train = MultiThreadedAugmenter( dataloader_train, tr_transforms, params.get('num_threads'), params.get("num_cached_per_thread"), seeds=seeds_train, pin_memory=pin_memory) val_transforms = [] val_transforms.append(RemoveLabelTransform(-1, 0)) if params.get("selected_data_channels") is not None: val_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) if params.get("selected_seg_channels") is not None: val_transforms.append( SegChannelSelectionTransform(params.get("selected_seg_channels"))) if params.get("move_last_seg_chanel_to_data") is not None and params.get( "move_last_seg_chanel_to_data"): val_transforms.append( MoveSegAsOneHotToData(1, params.get("all_segmentation_labels"), 'seg', 'data')) val_transforms.append(RenameTransform('seg', 'target', True)) if deep_supervision_scales is not None: if soft_ds: assert classes is not None val_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'target', 'target', classes)) else: val_transforms.append( DownsampleSegForDSTransform2(deep_supervision_scales, 0, 0, input_key='target', output_key='target')) val_transforms.append(NumpyToTensor(['data', 'target'], 'float')) val_transforms = Compose(val_transforms) batchgenerator_val = MultiThreadedAugmenter( dataloader_val, val_transforms, max(params.get('num_threads') // 2, 1), params.get("num_cached_per_thread"), seeds=seeds_val, pin_memory=pin_memory) return batchgenerator_train, batchgenerator_val
# logging # in order to use neptune logging: # export NEPTUNE_API_TOKEN = '...' !!! logging.getLogger().setLevel('INFO') source_files = [__file__] if hparams.config: source_files.append(hparams.config) neptune_logger = NeptuneLogger(project_name=hparams.neptune_project, params=vars(hparams), experiment_name=hparams.experiment_name, tags=[hparams.experiment_name], upload_source_files=source_files) tb_logger = loggers.TensorBoardLogger(hparams.log_dir) transform = Compose([ BrightnessTransform(mu=0.0, sigma=0.3, data_key='data'), GammaTransform(gamma_range=(0.7, 1.3), data_key='data'), ContrastAugmentationTransform(contrast_range=(0.3, 1.7), data_key='data') ]) with open(hparams.train_set, 'r') as keyfile: train_keys = [l.strip() for l in keyfile.readlines()] print(train_keys) with open(hparams.val_set, 'r') as keyfile: val_keys = [l.strip() for l in keyfile.readlines()] print(val_keys) train_ds = MedDataset(hparams.data_path, train_keys, hparams.patches_per_subject,
def get_insaneDA_augmentation(dataloader_train, dataloader_val, patch_size, params=default_3D_augmentation_params, border_val_seg=-1, seeds_train=None, seeds_val=None, order_seg=1, order_data=3, deep_supervision_scales=None, soft_ds=False, classes=None, pin_memory=True, regions=None): assert params.get( 'mirror') is None, "old version of params, use new keyword do_mirror" tr_transforms = [] # 'patch_size': array([288, 320]), # 'border_val_seg': -1, # 'seeds_train': None, # 'seeds_val': None, # 'order_seg': 1, # 'order_data': 3, # 'deep_supervision_scales': [[1, 1, 1], # [1.0, 0.5, 0.5], # [1.0, 0.25, 0.25], # [0.5, 0.125, 0.125], # [0.5, 0.0625, 0.0625]], # 'soft_ds': False, # 'classes': None, # 'pin_memory': True, # 'regions': None # params # {'selected_data_channels': None, # 'selected_seg_channels': [0], # 'do_elastic': True, # 'elastic_deform_alpha': (0.0, 300.0), # 'elastic_deform_sigma': (9.0, 15.0), # 'p_eldef': 0.1, # 'do_scaling': True, # 'scale_range': (0.65, 1.6), # 'independent_scale_factor_for_each_axis': True, # 'p_independent_scale_per_axis': 0.3, # 'p_scale': 0.3, # 'do_rotation': True, # 'rotation_x': (-3.141592653589793, 3.141592653589793), # 'rotation_y': (-0.5235987755982988, 0.5235987755982988), # 'rotation_z': (-0.5235987755982988, 0.5235987755982988), # 'rotation_p_per_axis': 1, # 'p_rot': 0.7, # 'random_crop': False, # 'random_crop_dist_to_border': None, # 'do_gamma': True, # 'gamma_retain_stats': True, # 'gamma_range': (0.5, 1.6), # 'p_gamma': 0.3, # 'do_mirror': True, # 'mirror_axes': (0, 1, 2), # 'dummy_2D': True, # 'mask_was_used_for_normalization': OrderedDict([(0, False)]), # 'border_mode_data': 'constant', # 'all_segmentation_labels': None, # 'move_last_seg_chanel_to_data': False, # 'cascade_do_cascade_augmentations': False, # 'cascade_random_binary_transform_p': 0.4, # 'cascade_random_binary_transform_p_per_label': 1, # 'cascade_random_binary_transform_size': (1, 8), # 'cascade_remove_conn_comp_p': 0.2, # 'cascade_remove_conn_comp_max_size_percent_threshold': 0.15, # 'cascade_remove_conn_comp_fill_with_other_class_p': 0.0, # 'do_additive_brightness': True, # 'additive_brightness_p_per_sample': 0.3, # 'additive_brightness_p_per_channel': 1, # 'additive_brightness_mu': 0, # 'additive_brightness_sigma': 0.2, # 'num_threads': 12, # 'num_cached_per_thread': 1, # 'patch_size_for_spatialtransform': array([288, 320])} # selected_data_channels is None if params.get("selected_data_channels") is not None: tr_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) # selected_seg_channels is [0] if params.get("selected_seg_channels") is not None: tr_transforms.append( SegChannelSelectionTransform(params.get("selected_seg_channels"))) # don't do color augmentations while in 2d mode with 3d data because the color channel is overloaded!! # dummy_2D is True if params.get("dummy_2D") is not None and params.get("dummy_2D"): ignore_axes = (0, ) tr_transforms.append(Convert3DTo2DTransform()) else: ignore_axes = None tr_transforms.append( SpatialTransform(patch_size, patch_center_dist_from_border=None, do_elastic_deform=params.get("do_elastic"), alpha=params.get("elastic_deform_alpha"), sigma=params.get("elastic_deform_sigma"), do_rotation=params.get("do_rotation"), angle_x=params.get("rotation_x"), angle_y=params.get("rotation_y"), angle_z=params.get("rotation_z"), do_scale=params.get("do_scaling"), scale=params.get("scale_range"), border_mode_data=params.get("border_mode_data"), border_cval_data=0, order_data=order_data, border_mode_seg="constant", border_cval_seg=border_val_seg, order_seg=order_seg, random_crop=params.get("random_crop"), p_el_per_sample=params.get("p_eldef"), p_scale_per_sample=params.get("p_scale"), p_rot_per_sample=params.get("p_rot"), independent_scale_for_each_axis=params.get( "independent_scale_factor_for_each_axis"), p_independent_scale_per_axis=params.get( "p_independent_scale_per_axis"))) if params.get("dummy_2D"): tr_transforms.append(Convert2DTo3DTransform()) # we need to put the color augmentations after the dummy 2d part (if applicable). Otherwise the overloaded color # channel gets in the way tr_transforms.append(GaussianNoiseTransform(p_per_sample=0.15)) tr_transforms.append( GaussianBlurTransform((0.5, 1.5), different_sigma_per_channel=True, p_per_sample=0.2, p_per_channel=0.5)) tr_transforms.append( BrightnessMultiplicativeTransform(multiplier_range=(0.70, 1.3), p_per_sample=0.15)) tr_transforms.append( ContrastAugmentationTransform(contrast_range=(0.65, 1.5), p_per_sample=0.15)) tr_transforms.append( SimulateLowResolutionTransform(zoom_range=(0.5, 1), per_channel=True, p_per_channel=0.5, order_downsample=0, order_upsample=3, p_per_sample=0.25, ignore_axes=ignore_axes)) tr_transforms.append( GammaTransform(params.get("gamma_range"), True, True, retain_stats=params.get("gamma_retain_stats"), p_per_sample=0.15)) # inverted gamma # do_additive_brightness is True if params.get("do_additive_brightness"): tr_transforms.append( BrightnessTransform( params.get("additive_brightness_mu"), params.get("additive_brightness_sigma"), True, p_per_sample=params.get("additive_brightness_p_per_sample"), p_per_channel=params.get("additive_brightness_p_per_channel"))) # do_gamma is True if params.get("do_gamma"): tr_transforms.append( GammaTransform(params.get("gamma_range"), False, True, retain_stats=params.get("gamma_retain_stats"), p_per_sample=params["p_gamma"])) # do_mirror is True if params.get("do_mirror") or params.get("mirror"): tr_transforms.append(MirrorTransform(params.get("mirror_axes"))) # mask_was_used_for_normalization is OrderedDict([(0, False)]), if params.get("mask_was_used_for_normalization") is not None: mask_was_used_for_normalization = params.get( "mask_was_used_for_normalization") tr_transforms.append( MaskTransform(mask_was_used_for_normalization, mask_idx_in_seg=0, set_outside_to=0)) tr_transforms.append(RemoveLabelTransform(-1, 0)) # move_last_seg_chanel_to_data is False if params.get("move_last_seg_chanel_to_data") is not None and params.get( "move_last_seg_chanel_to_data"): tr_transforms.append( MoveSegAsOneHotToData(1, params.get("all_segmentation_labels"), 'seg', 'data')) if params.get("cascade_do_cascade_augmentations" ) and not None and params.get( "cascade_do_cascade_augmentations"): if params.get("cascade_random_binary_transform_p") > 0: tr_transforms.append( ApplyRandomBinaryOperatorTransform( channel_idx=list( range(-len(params.get("all_segmentation_labels")), 0)), p_per_sample=params.get( "cascade_random_binary_transform_p"), key="data", strel_size=params.get( "cascade_random_binary_transform_size"))) if params.get("cascade_remove_conn_comp_p") > 0: tr_transforms.append( RemoveRandomConnectedComponentFromOneHotEncodingTransform( channel_idx=list( range(-len(params.get("all_segmentation_labels")), 0)), key="data", p_per_sample=params.get("cascade_remove_conn_comp_p"), fill_with_other_class_p=params.get( "cascade_remove_conn_comp_max_size_percent_threshold" ), dont_do_if_covers_more_than_X_percent=params.get( "cascade_remove_conn_comp_fill_with_other_class_p") )) tr_transforms.append(RenameTransform('seg', 'target', True)) # regions is None if regions is not None: tr_transforms.append( ConvertSegmentationToRegionsTransform(regions, 'target', 'target')) # deep_supervision_scales is a not None if deep_supervision_scales is not None: # soft_ds is False if soft_ds: assert classes is not None tr_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'target', 'target', classes)) else: tr_transforms.append( DownsampleSegForDSTransform2(deep_supervision_scales, 0, 0, input_key='target', output_key='target')) tr_transforms.append(NumpyToTensor(['data', 'target'], 'float')) tr_transforms = Compose(tr_transforms) batchgenerator_train = MultiThreadedAugmenter( dataloader_train, tr_transforms, params.get('num_threads'), params.get("num_cached_per_thread"), seeds=seeds_train, pin_memory=pin_memory) # ======================================================== val_transforms = [] val_transforms.append(RemoveLabelTransform(-1, 0)) # selected_data_channels is None if params.get("selected_data_channels") is not None: val_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) # selected_seg_channels is [0] if params.get("selected_seg_channels") is not None: val_transforms.append( SegChannelSelectionTransform(params.get("selected_seg_channels"))) # move_last_seg_chanel_to_data is False if params.get("move_last_seg_chanel_to_data") is not None and params.get( "move_last_seg_chanel_to_data"): val_transforms.append( MoveSegAsOneHotToData(1, params.get("all_segmentation_labels"), 'seg', 'data')) val_transforms.append(RenameTransform('seg', 'target', True)) # regions is None if regions is not None: val_transforms.append( ConvertSegmentationToRegionsTransform(regions, 'target', 'target')) # deep_supervision_scales is not None if deep_supervision_scales is not None: # soft_ds is False if soft_ds: assert classes is not None val_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'target', 'target', classes)) else: val_transforms.append( DownsampleSegForDSTransform2(deep_supervision_scales, 0, 0, input_key='target', output_key='target')) val_transforms.append(NumpyToTensor(['data', 'target'], 'float')) val_transforms = Compose(val_transforms) batchgenerator_val = MultiThreadedAugmenter( dataloader_val, val_transforms, max(params.get('num_threads') // 2, 1), params.get("num_cached_per_thread"), seeds=seeds_val, pin_memory=pin_memory) return batchgenerator_train, batchgenerator_val
def get_moreDA_augmentation( dataloader_train, dataloader_val, patch_size, params=default_3D_augmentation_params, border_val_seg=-1, seeds_train=None, seeds_val=None, order_seg=1, order_data=3, deep_supervision_scales=None, soft_ds=False, classes=None, pin_memory=True, anisotropy=False, extra_label_keys=None, val_mode=False, use_conf=False, ): ''' Work as Dataloader with augmentation :return: train_loader, val_loader for each iterator, return {'data': (B, D, H, W), 'target': (B, D, H, W)} ''' if not val_mode: assert params.get( 'mirror' ) is None, "old version of params, use new keyword do_mirror" tr_transforms = [] if params.get("selected_data_channels") is not None: tr_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) if params.get("selected_seg_channels") is not None: tr_transforms.append( SegChannelSelectionTransform( params.get("selected_seg_channels"))) # anistropic setting if anisotropy or params.get("dummy_2D"): ignore_axes = (0, ) tr_transforms.append( Convert3DTo2DTransform(extra_label_keys=extra_label_keys)) patch_size = patch_size[1:] # 2D patch size print('Using dummy2d data augmentation') params["elastic_deform_alpha"] = (0., 200.) params["elastic_deform_sigma"] = (9., 13.) params["rotation_x"] = (-180. / 360 * 2. * np.pi, 180. / 360 * 2. * np.pi) params["rotation_y"] = (-0. / 360 * 2. * np.pi, 0. / 360 * 2. * np.pi) params["rotation_z"] = (-0. / 360 * 2. * np.pi, 0. / 360 * 2. * np.pi) else: ignore_axes = None # 1. Spatial Transform: rotation, scaling tr_transforms.append( SpatialTransform(patch_size, patch_center_dist_from_border=None, do_elastic_deform=params.get("do_elastic"), alpha=params.get("elastic_deform_alpha"), sigma=params.get("elastic_deform_sigma"), do_rotation=params.get("do_rotation"), angle_x=params.get("rotation_x"), angle_y=params.get("rotation_y"), angle_z=params.get("rotation_z"), p_rot_per_axis=params.get("rotation_p_per_axis"), do_scale=params.get("do_scaling"), scale=params.get("scale_range"), border_mode_data=params.get("border_mode_data"), border_cval_data=0, order_data=order_data, border_mode_seg="constant", border_cval_seg=border_val_seg, order_seg=order_seg, random_crop=params.get("random_crop"), p_el_per_sample=params.get("p_eldef"), p_scale_per_sample=params.get("p_scale"), p_rot_per_sample=params.get("p_rot"), independent_scale_for_each_axis=params.get( "independent_scale_factor_for_each_axis"), extra_label_keys=extra_label_keys)) if anisotropy or params.get("dummy_2D"): tr_transforms.append( Convert2DTo3DTransform(extra_label_keys=extra_label_keys)) # 2. Noise Augmentation: gaussian noise, gaussian blur tr_transforms.append(GaussianNoiseTransform(p_per_sample=0.1)) tr_transforms.append( GaussianBlurTransform((0.5, 1.), different_sigma_per_channel=True, p_per_sample=0.2, p_per_channel=0.5)) # 3. Color Augmentation: brightness, constrast, low resolution, gamma_transform tr_transforms.append( BrightnessMultiplicativeTransform(multiplier_range=(0.75, 1.25), p_per_sample=0.15)) if params.get("do_additive_brightness"): tr_transforms.append( BrightnessTransform(params.get("additive_brightness_mu"), params.get("additive_brightness_sigma"), True, p_per_sample=params.get( "additive_brightness_p_per_sample"), p_per_channel=params.get( "additive_brightness_p_per_channel"))) tr_transforms.append(ContrastAugmentationTransform(p_per_sample=0.15)) tr_transforms.append( SimulateLowResolutionTransform(zoom_range=(0.5, 1), per_channel=True, p_per_channel=0.5, order_downsample=0, order_upsample=3, p_per_sample=0.25, ignore_axes=ignore_axes)) tr_transforms.append( GammaTransform(params.get("gamma_range"), True, True, retain_stats=params.get("gamma_retain_stats"), p_per_sample=0.1)) # inverted gamma if params.get("do_gamma"): tr_transforms.append( GammaTransform(params.get("gamma_range"), False, True, retain_stats=params.get("gamma_retain_stats"), p_per_sample=params["p_gamma"])) # 4. Mirror Transform if params.get("do_mirror") or params.get("mirror"): tr_transforms.append( MirrorTransform(params.get("mirror_axes"), extra_label_keys=extra_label_keys)) # if params.get("mask_was_used_for_normalization") is not None: # mask_was_used_for_normalization = params.get("mask_was_used_for_normalization") # tr_transforms.append(MaskTransform(mask_was_used_for_normalization, mask_idx_in_seg=0, set_outside_to=0)) tr_transforms.append( RemoveLabelTransform(-1, 0, extra_label_keys=extra_label_keys)) tr_transforms.append(RenameTransform('data', 'image', True)) tr_transforms.append(RenameTransform('seg', 'gt', True)) if deep_supervision_scales is not None: if soft_ds: assert classes is not None tr_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'gt', 'gt', classes)) else: tr_transforms.append( DownsampleSegForDSTransform2( deep_supervision_scales, 0, 0, input_key='gt', output_key='gt', extra_label_keys=extra_label_keys)) toTensorKeys = [ 'image', 'gt' ] + extra_label_keys if extra_label_keys is not None else [ 'image', 'gt' ] tr_transforms.append(NumpyToTensor(toTensorKeys, 'float')) tr_transforms = Compose(tr_transforms) if seeds_train is not None: seeds_train = [seeds_train] * params.get('num_threads') if use_conf: num_threads = 1 num_cached_per_thread = 1 else: num_threads, num_cached_per_thread = params.get( 'num_threads'), params.get("num_cached_per_thread") batchgenerator_train = MultiThreadedAugmenter(dataloader_train, tr_transforms, num_threads, num_cached_per_thread, seeds=seeds_train, pin_memory=pin_memory) val_transforms = [] val_transforms.append( RemoveLabelTransform(-1, 0, extra_label_keys=extra_label_keys)) if params.get("selected_data_channels") is not None: val_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) if params.get("selected_seg_channels") is not None: val_transforms.append( SegChannelSelectionTransform( params.get("selected_seg_channels"))) val_transforms.append(RenameTransform('data', 'image', True)) val_transforms.append(RenameTransform('seg', 'gt', True)) if deep_supervision_scales is not None: if soft_ds: assert classes is not None val_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'gt', 'gt', classes)) else: val_transforms.append( DownsampleSegForDSTransform2( deep_supervision_scales, 0, 0, input_key='gt', output_key='gt', extra_label_keys=extra_label_keys)) val_transforms.append(NumpyToTensor(toTensorKeys, 'float')) val_transforms = Compose(val_transforms) # batchgenerator_val = MultiThreadedAugmenter(dataloader_val, val_transforms, max(params.get('num_threads') // 2, 1), # params.get("num_cached_per_thread"), # seeds=seeds_val, pin_memory=pin_memory) if seeds_val is not None: seeds_val = [seeds_val] * 1 # batchgenerator_val = MultiThreadedAugmenter(dataloader_val, val_transforms, 1, # params.get("num_cached_per_thread"), # seeds=seeds_val, pin_memory=False) batchgenerator_val = SingleThreadedAugmenter(dataloader_val, val_transforms) else: val_transforms = [] val_transforms.append( RemoveLabelTransform(-1, 0, extra_label_keys=extra_label_keys)) if params.get("selected_data_channels") is not None: val_transforms.append( DataChannelSelectionTransform( params.get("selected_data_channels"))) if params.get("selected_seg_channels") is not None: val_transforms.append( SegChannelSelectionTransform( params.get("selected_seg_channels"))) val_transforms.append(RenameTransform('data', 'image', True)) val_transforms.append(RenameTransform('seg', 'gt', True)) if deep_supervision_scales is not None: if soft_ds: assert classes is not None val_transforms.append( DownsampleSegForDSTransform3(deep_supervision_scales, 'gt', 'gt', classes)) else: val_transforms.append( DownsampleSegForDSTransform2( deep_supervision_scales, 0, 0, input_key='gt', output_key='gt', extra_label_keys=extra_label_keys)) toTensorKeys = [ 'image', 'gt' ] + extra_label_keys if extra_label_keys is not None else [ 'image', 'gt' ] val_transforms.append(NumpyToTensor(toTensorKeys, 'float')) val_transforms = Compose(val_transforms) batchgenerator_val = SingleThreadedAugmenter(dataloader_val, val_transforms) if dataloader_train is not None: batchgenerator_train = SingleThreadedAugmenter( dataloader_train, val_transforms) else: batchgenerator_train = None return batchgenerator_train, batchgenerator_val
np.array(img.shape) // 2, do_elastic_deform=False, do_rotation=True, angle_z=(0, 2 * np.pi), # 旋转 do_scale=True, scale=(0.3, 3.), # 缩放 border_mode_data='constant', border_cval_data=0, order_data=1, random_crop=False) my_transforms.append(spatial_transform) GaussianNoise = GaussianNoiseTransform() # 高斯噪声 my_transforms.append(GaussianNoise) GaussianBlur = GaussianBlurTransform() # 高斯模糊 my_transforms.append(GaussianBlur) Brightness = BrightnessTransform(0, 0.2) # 亮度 my_transforms.append(Brightness) brightness_transform = ContrastAugmentationTransform( (0.3, 3.), preserve_range=True) # 对比度 my_transforms.append(brightness_transform) SimulateLowResolution = SimulateLowResolutionTransform() # 低分辨率 my_transforms.append(SimulateLowResolution) Gamma = GammaTransform() # 伽马增强 my_transforms.append(Gamma) mirror_transform = MirrorTransform(axes=(0, 1)) # 镜像 my_transforms.append(mirror_transform) all_transforms = Compose(my_transforms) multithreaded_generator = MultiThreadedAugmenter(batchgen, all_transforms, 1, 2) t = multithreaded_generator.next()
def main(): # assign global args global args args = parser.parse_args() # make a folder for the experiment general_folder_name = args.output_path try: os.mkdir(general_folder_name) except OSError: pass # create train, test split, return the indices, patients in test_split wont be seen during whole training train_idx, val_idx, test_idx = CreateTrainValTestSplit( HistoFile_path=args.input_file, num_splits=args.num_splits, num_test_folds=args.num_test_folds, num_val_folds=args.num_val_folds, seed=args.seed) IDs = train_idx + val_idx print('size of training set {}'.format(len(train_idx))) print('size of validation set {}'.format(len(val_idx))) print('size of test set {}'.format(len(test_idx))) # data loading Data = ProstataData(args.input_file) #For details on this class see README # train and validate for cv in range(args.cv_start, args.cv_end): best_epoch = 0 train_loss = [] val_loss = [] # define patients for training and validation train_idx, val_idx = split_training(IDs, len_val=62, cv=cv, cv_runs=args.cv_number) oversampling_factor, Slices_total, Natural_probability_tu_slice, Natural_probability_PRO_slice = get_oversampling( Data, train_idx=sorted(train_idx), Batch_Size=args.b, patch_size=args.patch_size) training_batches = Slices_total / args.b lr = args.lr base_lr = args.lr args.seed += 1 print('train_idx', train_idx, len(train_idx)) print('val_idx', val_idx, len(val_idx)) # get class frequencies print('calculating class frequencie') Tumor_frequencie_ADC, Prostate_frequencie_ADC, Background_frequencie_ADC,\ Tumor_frequencie_T2, Prostate_frequencie_T2, Background_frequencie_T2\ , ADC_mean, ADC_std, BVAL_mean, BVAL_std, T2_mean, T2_std \ = get_class_frequencies(Data, train_idx, patch_size=args.patch_size) print ADC_mean, ADC_std, BVAL_mean, BVAL_std, T2_mean, T2_std print('ADC', Tumor_frequencie_ADC, Prostate_frequencie_ADC, Background_frequencie_ADC) print('T2', Tumor_frequencie_T2, Prostate_frequencie_T2, Background_frequencie_T2) all_ADC = np.float(Background_frequencie_ADC + Prostate_frequencie_ADC + Tumor_frequencie_ADC) all_T2 = np.float(Background_frequencie_T2 + Prostate_frequencie_T2 + Tumor_frequencie_T2) print all_ADC print all_T2 W1_ADC = 1 / (Background_frequencie_ADC / all_ADC)**0.25 W2_ADC = 1 / (Prostate_frequencie_ADC / all_ADC)**0.25 W3_ADC = 1 / (Tumor_frequencie_ADC / all_ADC)**0.25 Wa_ADC = W1_ADC / (W1_ADC + W2_ADC + W3_ADC) Wb_ADC = W2_ADC / (W1_ADC + W2_ADC + W3_ADC) Wc_ADC = W3_ADC / (W1_ADC + W2_ADC + W3_ADC) print 'Weights ADC', Wa_ADC, Wb_ADC, Wc_ADC weight_ADC = (Wa_ADC, Wb_ADC, Wc_ADC) W1_T2 = 1 / (Background_frequencie_T2 / all_T2)**0.25 W2_T2 = 1 / (Prostate_frequencie_T2 / all_T2)**0.25 W3_T2 = 1 / (Tumor_frequencie_T2 / all_T2)**0.25 Wa_T2 = W1_T2 / (W1_T2 + W2_T2 + W3_T2) Wb_T2 = W2_T2 / (W1_T2 + W2_T2 + W3_T2) Wc_T2 = W3_T2 / (W1_T2 + W2_T2 + W3_T2) print 'Weights T2', Wa_T2, Wb_T2, Wc_T2 weight_T2 = (Wa_T2, Wb_T2, Wc_T2) # define model Net = UNetPytorch(in_shape=(3, args.patch_size[0], args.patch_size[1])) Net_Name = 'UNetPytorch' model = Net.cuda() # model parameter optimizer = torch.optim.Adam(model.parameters(), lr, weight_decay=args.weight_decay) criterion_ADC = CrossEntropyLoss2d( weight=torch.FloatTensor(weight_ADC)).cuda() criterion_T2 = CrossEntropyLoss2d( weight=torch.FloatTensor(weight_T2)).cuda() # new folder name for cv folder_name = general_folder_name + '/CV_{}'.format(cv) try: os.mkdir(folder_name) except OSError: pass checkpoint_file = folder_name + '/checkpoint_' + '{}.pth.tar'.format( Net_Name) # augmentation for epoch in range(args.epochs): torch.manual_seed(args.seed + epoch + cv) np.random.seed(epoch + cv) np.random.shuffle(train_idx) if epoch == 0: my_transforms = [] spatial_transform = SpatialTransform( args.patch_size, np.array(args.patch_size) // 2, do_elastic_deform=True, alpha=(100., 450.), sigma=(13., 17.), do_rotation=True, angle_z=(-np.pi / 2., np.pi / 2.), do_scale=True, scale=(0.75, 1.25), border_mode_data='constant', border_cval_data=0, order_data=3, random_crop=True) resample_transform = ResampleTransform(zoom_range=(0.7, 1.3)) brightness_transform = BrightnessTransform(0.0, 0.1, True) my_transforms.append(resample_transform) my_transforms.append( ContrastAugmentationTransform((0.75, 1.25), True)) my_transforms.append(brightness_transform) my_transforms.append(Mirror((2, 3))) all_transforms = Compose(my_transforms) sometimes_spatial_transforms = RndTransform( spatial_transform, prob=args.p, alternative_transform=CenterCropTransform(args.patch_size)) sometimes_other_transforms = RndTransform(all_transforms, prob=1.0) final_transform = Compose( [sometimes_spatial_transforms, sometimes_other_transforms]) Center_Crop = CenterCropTransform(args.patch_size) if epoch == 30: my_transforms = [] spatial_transform = SpatialTransform( args.patch_size, np.array(args.patch_size) // 2, do_elastic_deform=True, alpha=(0., 250.), sigma=(11., 14.), do_rotation=True, angle_z=(-np.pi / 2., np.pi / 2.), do_scale=True, scale=(0.85, 1.15), border_mode_data='constant', border_cval_data=0, order_data=3, random_crop=True) resample_transform = ResampleTransform(zoom_range=(0.8, 1.2)) brightness_transform = BrightnessTransform(0.0, 0.1, True) my_transforms.append(resample_transform) my_transforms.append( ContrastAugmentationTransform((0.85, 1.15), True)) my_transforms.append(brightness_transform) all_transforms = Compose(my_transforms) sometimes_spatial_transforms = RndTransform( spatial_transform, prob=args.p, alternative_transform=CenterCropTransform(args.patch_size)) sometimes_other_transforms = RndTransform(all_transforms, prob=1.0) final_transform = Compose( [sometimes_spatial_transforms, sometimes_other_transforms]) Center_Crop = CenterCropTransform(args.patch_size) if epoch == 50: my_transforms = [] spatial_transform = SpatialTransform( args.patch_size, np.array(args.patch_size) // 2, do_elastic_deform=True, alpha=(0., 150.), sigma=(10., 12.), do_rotation=True, angle_z=(-np.pi / 2., np.pi / 2.), do_scale=True, scale=(0.85, 1.15), border_mode_data='constant', border_cval_data=0, order_data=3, random_crop=False) resample_transform = ResampleTransform(zoom_range=(0.9, 1.1)) brightness_transform = BrightnessTransform(0.0, 0.1, True) my_transforms.append(resample_transform) my_transforms.append( ContrastAugmentationTransform((0.95, 1.05), True)) my_transforms.append(brightness_transform) all_transforms = Compose(my_transforms) sometimes_spatial_transforms = RndTransform( spatial_transform, prob=args.p, alternative_transform=CenterCropTransform(args.patch_size)) sometimes_other_transforms = RndTransform(all_transforms, prob=1.0) final_transform = Compose( [sometimes_spatial_transforms, sometimes_other_transforms]) Center_Crop = CenterCropTransform(args.patch_size) train_loader = BatchGenerator( Data, BATCH_SIZE=args.b, split_idx=train_idx, seed=args.seed, ProbabilityTumorSlices=oversampling_factor, epoch=epoch, ADC_mean=ADC_mean, ADC_std=ADC_std, BVAL_mean=BVAL_mean, BVAL_std=BVAL_std, T2_mean=T2_mean, T2_std=T2_std) val_loader = BatchGenerator(Data, BATCH_SIZE=0, split_idx=val_idx, seed=args.seed, ProbabilityTumorSlices=None, epoch=epoch, test=True, ADC_mean=ADC_mean, ADC_std=ADC_std, BVAL_mean=BVAL_mean, BVAL_std=BVAL_std, T2_mean=T2_mean, T2_std=T2_std) #train on training set train_losses = train(train_loader=train_loader, model=model, optimizer=optimizer, criterion_ADC=criterion_ADC, criterion_T2=criterion_T2, final_transform=final_transform, workers=args.workers, seed=args.seed, training_batches=training_batches) train_loss.append(train_losses) # evaluate on validation set val_losses = validate(val_loader=val_loader, model=model, folder_name=folder_name, criterion_ADC=criterion_ADC, criterion_T2=criterion_T2, split_ixs=val_idx, epoch=epoch, workers=1, Center_Crop=Center_Crop, seed=args.seed) val_loss.append(val_losses) # write TrainingsCSV to folder name TrainingsCSV = pd.DataFrame({ 'train_loss': train_loss, 'val_loss': val_loss }) TrainingsCSV.to_csv(folder_name + '/TrainingsCSV.csv') if val_losses <= min(val_loss): best_epoch = epoch print 'best epoch', epoch save_checkpoint( { 'epoch': epoch, 'arch': args.arch, 'state_dict': model.state_dict(), 'optimizer': optimizer.state_dict() }, filename=checkpoint_file) optimizer, lr = adjust_learning_rate(optimizer, base_lr, epoch) # delete all output except best epoch clear_image_data(folder_name, best_epoch, epoch)