def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'unaligned_attr': from data.unaligned_attr_dataset import UnalignedAttrDataset dataset = UnalignedAttrDataset() elif opt.dataset_mode == 'unaligned_prog': from data.unaligned_prog_dataset import UnalignedProgDataset dataset = UnalignedProgDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'triple': from data.triple_dataset import TripleDataset dataset = TripleDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): ''' Gets called by CustomDatasetDataLoader.initialize(). dataset_mode is by default unaligned. Dataset has generic structure, inputs are coming from opts. Aligned, Unaligned are for A->B (i.e., image-to-image transfer type problems, whereas Single is for z->A problems (and testing). ''' dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'slice': from data.slice_dataset import SliceDataset dataset = SliceDataset() elif opt.dataset_mode == 'voxel': from data.voxel_dataset import VoxelDataset dataset = VoxelDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'star' or opt.dataset_mode == 'n2': from data.half_dataset import N2Dataset dataset = N2Dataset() elif opt.dataset_mode == 'star1' or opt.dataset_mode == 'n2t': from data.half_dataset import N2TestDataset dataset = N2TestDataset() elif opt.dataset_mode == 'star2' or opt.dataset_mode == '2n': from data.half_dataset import _2NDataset dataset = _2NDataset() elif opt.dataset_mode == 'single_star' or opt.dataset_mode == 'sn2t': from data.single_dataset import SingleN2Dataset dataset = SingleN2Dataset() elif opt.dataset_mode == 'single_star1' or opt.dataset_mode == 'snt': from data.single_dataset import SingleNDataset dataset = SingleNDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'yh': from data.yh_dataset import yhDataset dataset = yhDataset() elif opt.dataset_mode == 'yh_seg': from data.yh_seg_dataset import yhSegDataset dataset = yhSegDataset() elif opt.dataset_mode == 'yh_seg_spleen': from data.yh_seg_spleenonly_dataset import yhSegDatasetSpleenOnly dataset = yhSegDatasetSpleenOnly() elif opt.dataset_mode == 'yh_test_seg': from data.yh_test_seg_dataset import yhTestSegDataset dataset = yhTestSegDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'mat': from data.mat_dataset import MatDataset dataset = MatDataset() elif opt.dataset_mode == 'singlemat': from data.single_mat_dataset import SingleMatDataset dataset = SingleMatDataset() elif opt.dataset_mode == 'superpix': from data.superpix_dataset import SuperPixDataset dataset = SuperPixDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'unaligned_A_labeled': from data.unaligned_A_labeled_dataset import UnalignedALabeledDataset dataset = UnalignedALabeledDataset() elif opt.dataset_mode == 'EEG': from data.eeg_dataset import EEGDataset dataset = EEGDataset() elif opt.dataset_mode == 'EEGsingle': from data.eeg_single_dataset import EEGDataset dataset = EEGDataset() elif opt.dataset_mode == 'TestEEG': from data.eeg_dataset_test import EEGDataset dataset = EEGDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'thermal': from data.thermal_dataset import ThermalDataset dataset = ThermalDataset() elif opt.dataset_mode == 'thermal_rel': from data.thermal_rel_dataset import ThermalRelDataset dataset = ThermalRelDataset() elif opt.dataset_mode == 'fruxel': from data.fruxel_dataset import FruxelDataset dataset = FruxelDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'unaligned_random_crop': from data.unaligned_random_crop import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'pair': from data.pair_dataset import PairDataset dataset = PairDataset() elif opt.dataset_mode == 'syn': from data.syn_dataset import PairDataset dataset = PairDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'alignedrandom': from data.aligned_random_dataset import AlignedRandomDataset dataset = AlignedRandomDataset() elif opt.dataset_mode == 'Coco': from data.coco_dataset import UnalignedCocoDataset dataset = UnalignedCocoDataset() elif opt.dataset_mode == 'CocoSeg': from data.cocoseg_dataset import CocoSegDataset dataset = CocoSegDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'tif': from data.tif_dataset import TifDataset dataset = TifDataset(opt) elif opt.dataset_mode == 'mb': from data.mb_dataset import MBDataset dataset = MBDataset(opt) elif opt.dataset_mode == 'png_withlist': from data.png_dataset_withlist import PngDataset dataset = PngDataset(opt) elif opt.dataset_mode == 'png': from data.png_dataset import PngDataset dataset = PngDataset(opt) else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'unaligned_A_labeled': from data.unaligned_A_labeled_dataset import UnalignedALabeledDataset dataset = UnalignedALabeledDataset() elif opt.dataset_mode == 'mnist_svhn': from data.mnist_svhn_dataset import MnistSvhnDataset dataset = MnistSvhnDataset() elif opt.dataset_mode == 'svhn_mnist': from data.svhn_mnist_dataset import SvhnMnistDataset dataset = SvhnMnistDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'imagelist': from data.imagelist_dataset import ImageList if opt.isTrain: dataset = ImageList(root=opt.image_root, fileList=opt.train_list) else: dataset = ImageList(root=opt.image_root, fileList=opt.train_list, testPahse=True) elif opt.dataset_mode == 'imagelist_cross_view': from data.imagelist_dataset import ImageList_cross_view dataset = ImageList_cross_view() elif opt.dataset_mode == 'imglist_pts': from data.imagelist_pts_dataset import Imglist_Pts_Dataset dataset = Imglist_Pts_Dataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def ISTD_test(self): opt = self.opt opt.mask_test = '/home/balin/exper/shadow_removal/Dataset/ISTD_Dataset/test/test_B' dataset = SingleDataset( '/home/balin/exper/shadow_removal/Dataset/ISTD_Dataset/test/test_A', opt) opt.results_dir = './ISTD_b/' self.eval_backend_output_only(dataset, opt)
def CreateSingleDataset(opt): dataset = None from data.single_dataset import SingleDataset dataset = SingleDataset() print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def create_dataset(args): if args.phase == 'train': rgb_transforms, torchvision_transforms, gt_transforms = get_transform( args) return AlignedDataset(args.data_root, rgb_transforms, torchvision_transforms, gt_transforms) elif args.phase == 'test': rgb_transforms = get_transform(args) return SingleDataset(args.data_root, rgb_transforms)
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'aligned_with_C': from data.aligned_dataset_with_C import AlignedDatasetWithC dataset = AlignedDatasetWithC() elif opt.dataset_mode == 'aligned_multi_view': from data.aligned_dataset_multi_view import AlignedDatasetMultiView dataset = AlignedDatasetMultiView() elif opt.dataset_mode == 'aligned_multi_view_random': from data.aligned_dataset_multi_view_random import AlignedDatasetMultiView dataset = AlignedDatasetMultiView() elif opt.dataset_mode == 'aligned_depth': from data.aligned_dataset_depth import AlignedDatasetDepth dataset = AlignedDatasetDepth() elif opt.dataset_mode == 'appearance_flow': from data.appearance_flow_dataloader import AppearanceFlowDataloader dataset = AppearanceFlowDataloader() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'unaligned_with_guidance': from data.unaligned_dataset_with_guidance import UnalignedDatasetWithGuidance dataset = UnalignedDatasetWithGuidance() elif opt.dataset_mode == 'unaligned_with_label': from data.unaligned_dataset_with_label import UnalignedDatasetWithLabel dataset = UnalignedDatasetWithLabel() elif opt.dataset_mode == 'unaligned_tensor_with_label': from data.unaligned_tensor_dataset_with_label import UnalignedTensorDatasetWithLabel dataset = UnalignedTensorDatasetWithLabel() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None # Data stored as one image concatenated along horizontal axis if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() # Data stored in different directories elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'geo': from data.geo_dataset import GeoDataset dataset = GeoDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None #print("================="+str(opt.dataset_mode)) if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() ## add 3D videodataset loader elif opt.dataset_mode == 'v': from data.video_data import VideoDataset dataset = VideoDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'unaligned_landmark': from data.unaligned_landmark_dataset import UnalignedLandmarkDataset dataset = UnalignedLandmarkDataset() elif opt.dataset_mode == 'aligned_heatmap2face': from data.aligned_dataset import AlignedDatasetHeatmap2Face dataset = AlignedDatasetHeatmap2Face() elif opt.dataset_mode == 'aligned_boundary_detection': from data.aligned_dataset import AlignedBoundaryDetection dataset = AlignedBoundaryDetection() elif opt.dataset_mode == 'aligned_boundary_detection_landmarks': from data.aligned_dataset import AlignedBoundaryDetectionLandmark dataset = AlignedBoundaryDetectionLandmark() elif opt.dataset_mode == 'aligned_face2boundary2face': from data.aligned_dataset import AlignedFace2Boudnary2Face dataset = AlignedFace2Boudnary2Face() elif opt.dataset_mode == 'aligned_face2face': from data.aligned_dataset import AlignedFace2Face dataset = AlignedFace2Face() elif opt.dataset_mode == 'aligned_faceDataset': from data.aligned_dataset import AlignedFaceDataset dataset = AlignedFaceDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'aligned_rand': from data.aligned_dataset_rand import AlignedDataset_Rand dataset = AlignedDataset_Rand() elif opt.dataset_mode == 'aligned_test': from data.aligned_dataset_test import AlignedDataset_Test dataset = AlignedDataset_Test() elif opt.dataset_mode == 'unaligned_seg': from data.unaligned_dataset_seg import UnalignedDataset_Seg dataset = UnalignedDataset_Seg() elif opt.dataset_mode == 'aligned_seg': from data.aligned_dataset_seg import AlignedDataset_Seg dataset = AlignedDataset_Seg() elif opt.dataset_mode == 'aligned_seg_rand': from data.aligned_dataset_seg_rand import AlignedDataset_Seg_Rand dataset = AlignedDataset_Seg_Rand() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
opt.no_flip = True # no flip; comment this line if results on flipped images are needed. opt.display_id = -1 # no visdom display; the test code saves the results to a HTML file. opt.direction = 'AtoB' opt.gpu_ids = [0] opt.netG = 'unet_256' opt.norm = 'batch' opt.epoch = 15 opt.num_test = float("inf") model = Pix2PixModel(opt) model.save_dir = './vhs2film' epoch = opt.epoch model.load_networks(epoch) model.eval() dataset = SingleDataset(opt) dataloader = torch.utils.data.DataLoader(dataset, batch_size=opt.batch_size, shuffle=False, num_workers=1) for i, data in enumerate(dataloader): # print(data) if i >= opt.num_test: # only apply our model to opt.num_test images. break model.set_input(data) # unpack data from data loader with torch.no_grad(): model.forward() # run inference image_numpy = util.tensor2im(model.fake_B) im = Image.fromarray(image_numpy) png_i = "out" + str(opt.epoch) + "/%06d.png" % i
def CreateDataset(opt): dataset = None if opt.dataset_mode == 'aligned': from data.aligned_dataset import AlignedDataset dataset = AlignedDataset() elif opt.dataset_mode == 'unaligned': from data.unaligned_dataset import UnalignedDataset dataset = UnalignedDataset() elif opt.dataset_mode == 'aligned_rand': from data.aligned_dataset_rand import AlignedDataset_Rand dataset = AlignedDataset_Rand() elif opt.dataset_mode == 'aligned_test': from data.aligned_dataset_test import AlignedDataset_Test dataset = AlignedDataset_Test() elif opt.dataset_mode == 'unaligned_seg': from data.unaligned_dataset_seg import UnalignedDataset_Seg dataset = UnalignedDataset_Seg() elif opt.dataset_mode == 'aligned_seg': from data.aligned_dataset_seg import AlignedDataset_Seg dataset = AlignedDataset_Seg() elif opt.dataset_mode == 'aligned_seg_rand': from data.aligned_dataset_seg_rand import AlignedDataset_Seg_Rand dataset = AlignedDataset_Seg_Rand() elif opt.dataset_mode == 'single': from data.single_dataset import SingleDataset dataset = SingleDataset() elif opt.dataset_mode == 'fivek': from data.fivek_dataset import FiveKDataset dataset = FiveKDataset() elif opt.dataset_mode == 'fivek2': from data.fivek_dataset2 import FiveKDataset2 dataset = FiveKDataset2() elif opt.dataset_mode == 'fivek3': from data.fivek_dataset3 import FiveKDataset3 dataset = FiveKDataset3() elif opt.dataset_mode == 'fivek4': from data.fivek_dataset4 import FiveKDataset4 dataset = FiveKDataset4() elif opt.dataset_mode == 'fivek4_syn': from data.fivek_dataset4_syn import FiveKDataset4_syn dataset = FiveKDataset4_syn() elif opt.dataset_mode == 'fivek_single': from data.fivek_single import FiveKDataset_single dataset = FiveKDataset_single() elif opt.dataset_mode == 'ava': from data.ava_dataset import AVADataset dataset = AVADataset() elif opt.dataset_mode == 'aadb': from data.aadb_dataset import AADBDataset dataset = AADBDataset() else: raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode) print("dataset [%s] was created" % (dataset.name())) dataset.initialize(opt) return dataset
def ISTD_test(self): opt = self.opt opt.mask_test = '/nfs/bigneuron/add_disk0/hieule/data/datasets/ISTD_Dataset/Mean_Teacher_SD/ISTD_crf' dataset = SingleDataset('/nfs/bigneuron/add_disk0/hieule/data/datasets/ISTD_Dataset/test/test_A',opt) opt.results_dir ='./ISTD_b/' self.eval_backend_output_only(dataset,opt)
from unet256 import Unet256Model from data.single_dataset import SingleDataset import numpy as np from PIL import Image import torch import tensorflow as tf from onnx_tf.backend import prepare import onnx from torch.autograd import Variable dataset = SingleDataset('../datasets/SBUsd/Test/TestA/') _data = None for i, data in enumerate(dataset): _data = data break import onnx import caffe2.python.onnx.backend as onnx_caffe2_backend model = Unet256Model(load_model='135_net_D.pth') model.print_net() # Export the model torch_out = torch.onnx.export( model.net, # model being run Variable(data['A'].cuda(0), requires_grad=0), # model input (or a tuple for multiple inputs) "model.onnx", # where to save the model (can be a file or file-like object) input_names=['input'], output_names=['output'])
from unet256 import Unet256Model from data.single_dataset import SingleDataset import numpy as np from PIL import Image dataset = SingleDataset('Test') print 'dataset size: ' + str(len(dataset)) model = Unet256Model(load_model='50_net_D.pth') model.print_net() for i, data in enumerate(dataset): print data['imname'] out = model.test(data) im_out = out[0].cpu().float().numpy() im_out = np.transpose(im_out, (1, 2, 0)) im_out = (im_out + 1) / 2 * np.log(256) im_out = np.exp(im_out) - 1 im_out = im_out.astype('uint8') Image.fromarray(np.squeeze(im_out, axis=2)).save('out/' + data['imname'])