def build_clevr_mv_dataset( args, configs, image_root, depth_root, scenes_json, questions_json ): import jactorch.transforms.bbox as T image_transform = T.Compose( [ # T.NormalizeBbox(), # T.Resize(configs.data.image_size), # T.DenormalizeBbox(), T.ToTensor(), T.Lambda(lambda x, y: (x - 0.5, y)) # T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ] ) from nscl.datasets.datasets import NSCLMultiviewDataset dataset = NSCLMultiviewDataset( scenes_json, questions_json, image_root=image_root, image_transform=image_transform, depth_root=depth_root, vocab_json=args.data_vocab_json, ) return dataset
def build_concept_quantization_clevr_dataset(args, configs, image_root, scenes_json): import jactorch.transforms.bbox as T image_transform = T.Compose([ T.NormalizeBbox(), T.Resize(configs.data.image_size), T.DenormalizeBbox(), T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) ]) from nscl.datasets.datasets import ConceptQuantizationDataset dataset = ConceptQuantizationDataset(scenes_json, image_root=image_root, image_transform=image_transform) return dataset
def build_clevr_dataset(args, configs, image_root, scenes_json, questions_json): import jactorch.transforms.bbox as T image_transform = T.Compose([ T.NormalizeBbox(), T.Resize(configs.data.image_size), T.DenormalizeBbox(), T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) ]) from nscl.datasets.datasets import NSCLDataset dataset = NSCLDataset( scenes_json, questions_json, image_root=image_root, image_transform=image_transform, vocab_json=args.data_vocab_json ) return dataset
def __init__(self, args, phase): self.args = args self.phase = phase self.loader = default_loader self.data_dir = args.data_dir self.label_dir = args.label_dir self.prp_dir = args.prp_dir self.ann_dir = args.ann_dir self.tube_dir = args.tube_dir self.valid_idx_lst = 'valid_idx_' + self.phase + '.txt' self.H = 100 self.W = 150 self.bbox_size = 24 ratio = self.args.train_valid_ratio n_train = round(self.args.n_rollout * ratio) if phase == 'train': self.st_idx = 0 self.n_rollout = n_train elif phase == 'valid': self.st_idx = n_train self.n_rollout = self.args.n_rollout - n_train else: raise AssertionError("Unknown phase") if self.args.gen_valid_idx: if self.args.visualize_flag == 1: self.gen_predict_input() elif self.args.version == 'v3': self.gen_valid_idx_from_tube_info_v3() else: self.gen_valid_idx_from_tube_info() else: self.read_valid_idx() self.img_transform = T.Compose([ T.Resize(self.args.img_size), T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) ]) self.W = 480 self.H = 320
def build_concept_retrieval_clevrer_dataset(args, configs, program, image_root, scenes_json): import jactorch.transforms.bbox as T image_transform = T.Compose([ T.NormalizeBbox(), T.Resize(configs.data.image_size), T.DenormalizeBbox(), T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ]) from nscl.datasets.datasets import ConceptRetrievalDataset dataset = ConceptRetrievalDataset(program, scenes_json, image_root=image_root, image_transform=image_transform) return dataset
def __init__(self, scenes_json, image_root, depth_root, incl_scene=True): super().__init__() self.scenes_json = scenes_json self.image_root = image_root self.depth_root = depth_root self.incl_scene = incl_scene logger.info('Loading scenes from: "{}".'.format(self.scenes_json)) self.scenes = io.load_json(self.scenes_json)["scenes"] import jactorch.transforms.bbox as T self.image_transform = T.Compose( [ # T.NormalizeBbox(), # T.Resize(configs.data.image_size), # T.DenormalizeBbox(), T.ToTensor(), T.Lambda(lambda x, y: (x - 0.5, y)) # T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ] )