def __init__(self, opt, data_num=None):
self.opt = opt
self.annotation_path = opt.infer_dataset_path
self.isTrain = False
self.data_list = data_utils.load_from_annotation(self.annotation_path)
# pad data list so that the number of data can be divisible by batch size
# the redundant data will be removed after the whole dataset has been processed
add_num = opt.batchSize - len(self.data_list)%opt.batchSize
self.data_list += self.data_list[:add_num]
# transform list
transform_list = [transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5),
(0.5, 0.5, 0.5))]
self.transform = transforms.Compose(transform_list)
self.data_processor = DataProcessor(opt)
def __init__(self, opt, data_num=0):
self.opt = opt
self.annotation_path = opt.human36m_anno_path
self.isTrain = opt.isTrain
data_list = self.load_annotation(self.annotation_path)
data_list = sorted(data_list, key=lambda a:a['image_path'])
self.update_path(opt.data_root, data_list)
if opt.isTrain and data_num > 0:
data_num = min(data_num, len(data_list))
data_list = random.sample(data_list, data_num)
self.data_list = data_list
transform_list = [ transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5),
(0.5, 0.5, 0.5))]
self.transform = transforms.Compose(transform_list)
self.data_processor = DataProcessor(opt)
def __init__(self, opt):
self.opt = opt
self.annotation_path = opt.up3d_anno_path
self.isTrain = opt.isTrain
self.refine_IUV = opt.refine_IUV
self.dp_num_max = opt.dp_num_max
data_list = self.load_annotation(self.annotation_path)
data_list = sorted(data_list, key=lambda a: a['image_path'])
self.update_path(opt.data_root, data_list)
self.data_list = data_list
if not opt.isTrain:
add_num = opt.batchSize - len(self.data_list) % opt.batchSize
self.data_list += self.data_list[:add_num]
transform_list = [
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
]
self.transform = transforms.Compose(transform_list)
self.data_processor = DataProcessor(opt)