dataloader = torch.utils.data.DataLoader(dataset, batch_size=args.batch_size, sampler=sampler_batch, num_workers=args.num_workers) # initilize the tensor holder here. im_left_data = Variable(torch.FloatTensor(1).cuda()) im_right_data = Variable(torch.FloatTensor(1).cuda()) im_info = Variable(torch.FloatTensor(1).cuda()) num_boxes = Variable(torch.LongTensor(1).cuda()) gt_boxes_left = Variable(torch.FloatTensor(1).cuda()) gt_boxes_right = Variable(torch.FloatTensor(1).cuda()) gt_boxes_merge = Variable(torch.FloatTensor(1).cuda()) gt_dim_orien = Variable(torch.FloatTensor(1).cuda()) gt_kpts = Variable(torch.FloatTensor(1).cuda()) # initilize the network here. stereoRCNN = resnet(imdb.classes, 101, pretrained=True) stereoRCNN.create_architecture() lr = cfg.TRAIN.LEARNING_RATE uncert = Variable(torch.rand(6).cuda(), requires_grad=True) torch.nn.init.constant(uncert, -1.0) params = [] for key, value in dict(stereoRCNN.named_parameters()).items(): if value.requires_grad: if 'bias' in key: params += [{'params':[value],'lr':lr*(cfg.TRAIN.DOUBLE_BIAS + 1), \ 'weight_decay': cfg.TRAIN.BIAS_DECAY and cfg.TRAIN.WEIGHT_DECAY or 0}] else:
args = parse_args() np.random.seed(cfg.RNG_SEED) input_dir = args.load_dir + "/" if not os.path.exists(input_dir): raise Exception( 'There is no input directory for loading network from ' + input_dir) load_name = os.path.join( input_dir, 'stereo_rcnn_{}_{}.pth'.format(args.checkepoch, args.checkpoint)) kitti_classes = np.asarray(['__background__', 'Car']) # initilize the network here. stereoRCNN = resnet(kitti_classes, 101, pretrained=False) stereoRCNN.create_architecture() print("load checkpoint %s" % (load_name)) checkpoint = torch.load(load_name) stereoRCNN.load_state_dict(checkpoint['model']) print('load model successfully!') with torch.no_grad(): # initilize the tensor holder here. im_left_data = Variable(torch.FloatTensor(1).cuda()) im_right_data = Variable(torch.FloatTensor(1).cuda()) im_info = Variable(torch.FloatTensor(1).cuda()) num_boxes = Variable(torch.LongTensor(1).cuda()) gt_boxes = Variable(torch.FloatTensor(1).cuda())