Next

natanielruiz

2017-10-30 2c764e41e2fde6244b87da58d12c40d09a14fcb4

 code/test_preangles.py

@@ -36,6 +36,13 @@

    return args

def load_filtered_state_dict(model, snapshot):
    # By user apaszke from discuss.pytorch.org
    model_dict = model.state_dict()
    snapshot = {k: v for k, v in snapshot.items() if k in model_dict}
    model_dict.update(snapshot)
    model.load_state_dict(model_dict)

if __name__ == '__main__':
    args = parse_args()

@@ -49,7 +56,7 @@
    print 'Loading snapshot.'
    # Load snapshot
    saved_state_dict = torch.load(snapshot_path)
    model.load_state_dict(saved_state_dict)
    load_filtered_state_dict(model, saved_state_dict)

    print 'Loading data.'

@@ -63,6 +70,8 @@
        pose_dataset = datasets.Pose_300W_LP_random_ds(args.data_dir, args.filename_list, transformations)
    elif args.dataset == 'AFLW2000':
        pose_dataset = datasets.AFLW2000(args.data_dir, args.filename_list, transformations)
    elif args.dataset == 'AFLW2000_ds':
        pose_dataset = datasets.AFLW2000_ds(args.data_dir, args.filename_list, transformations)
    elif args.dataset == 'BIWI':
        pose_dataset = datasets.BIWI(args.data_dir, args.filename_list, transformations)
    elif args.dataset == 'AFLW':
@@ -108,21 +117,16 @@
        _, roll_bpred = torch.max(roll.data, 1)

        # Continuous predictions
        yaw_predicted = utils.softmax_temperature(yaw.data, 1)
        pitch_predicted = utils.softmax_temperature(pitch.data, 1)
        roll_predicted = utils.softmax_temperature(roll.data, 1)

        yaw_predicted = torch.sum(yaw_predicted * idx_tensor, 1).cpu() * 3 - 99
        pitch_predicted = torch.sum(pitch_predicted * idx_tensor, 1).cpu() * 3 - 99
        roll_predicted = torch.sum(roll_predicted * idx_tensor, 1).cpu() * 3 - 99
        yaw_predicted = angles[:,0].data.cpu()
        pitch_predicted = angles[:,1].data.cpu()
        roll_predicted = angles[:,2].data.cpu()

        # Mean absolute error
        yaw_error += torch.sum(torch.abs(yaw_predicted - label_yaw))
        pitch_error += torch.sum(torch.abs(pitch_predicted - label_pitch))
        roll_error += torch.sum(torch.abs(roll_predicted - label_roll))

        # Save images with pose cube.
        # TODO: fix for larger batch size
        # Save first image in batch with pose cube or axis.
        if args.save_viz:
            name = name[0]
            if args.dataset == 'BIWI':