epoch_loss_dict[phase]['IoU'].append(epoch_IoU)
epoch_loss_dict[phase]['time'].append(t)
if detailed_time:
epoch_loss_dict[phase]['backward_pass_time'].append(total_pass)
epoch_loss_dict[phase]['data_fetch_time'].append(
total_data_fetch)
print("PHASE={} EPOCH={} TIME={} LOSS={} ACC={}".format(
phase, epoch, t, epoch_loss, epoch_acc))
return model, best_model_wts, epoch_loss_dict, batch_loss_dict
# Define trasnforms
common_transforms = [
transform_utils.RandomHorizontalFlip(0.5),
transform_utils.RandomVerticalFlip(0.5)
]
#img_transforms = [transforms.ColorJitter()]
# Define network
net = context_models.FrontEnd_ContextModel(FRONT_END_TYPE,
PATH_TO_FRONT_END_WEIGHTS, IS_GPU,
input_channels, img_size,
CONTEXT_LAYER_COUNT,
OUTPUT_CHANNELS)
net.fix_front_end_weights()
#net.load_vgg_weights(VGG_TRAIN)
# Define dataloaders
train_root = os.path.join(data_root, "train")
epoch_loss_dict[phase]['acc'].append(epoch_acc)
epoch_loss_dict[phase]['loss'].append(epoch_loss)
epoch_loss_dict[phase]['IoU'].append(epoch_IoU)
epoch_loss_dict[phase]['time'].append(t)
if detailed_time:
epoch_loss_dict[phase]['backward_pass_time'].append(total_pass)
epoch_loss_dict[phase]['data_fetch_time'].append(total_data_fetch)
print("PHASE={} EPOCH={} TIME={} LOSS={} ACC={}".format(phase,
epoch, t, epoch_loss, epoch_acc))
return model, best_model_wts, epoch_loss_dict, batch_loss_dict
# Define trasnforms
common_transforms = [transform_utils.RandomHorizontalFlip(0.5),
transform_utils.RandomVerticalFlip(0.5)]
#img_transforms = [transforms.ColorJitter()]
# Define network
net = context_models.FrontEnd_ContextModel(FRONT_END_TYPE, PATH_TO_FRONT_END_WEIGHTS, IS_GPU,
input_channels, img_size, CONTEXT_LAYER_COUNT, OUTPUT_CHANNELS)
net.fix_front_end_weights()
# Define dataloaders
train_root = os.path.join(data_root, "train")
val_root = os.path.join(data_root, "val")
train_dset = dataset_def.SegmentationDataset(train_root, list_common_trans=common_transforms,
list_img_trans=None, f_type = "PIL")
val_dset = dataset_def.SegmentationDataset(val_root, f_type="PIL")
