Function inference

trainer.py:22–38

(args, model, best_performance)

from utils.utils import powerset, one_hot_encoder, DiceLoss, val_single_volume
            
def inference(args, model, best_performance):
    db_test = Synapse_dataset(base_dir=args.volume_path, split="test_vol", list_dir=args.list_dir, nclass=args.num_classes)
    
    testloader = DataLoader(db_test, batch_size=1, shuffle=False, num_workers=1)
    logging.info("{} test iterations per epoch".format(len(testloader)))
    model.eval()
    metric_list = 0.0
    for i_batch, sampled_batch in tqdm(enumerate(testloader)):
        h, w = sampled_batch["image"].size()[2:]
        image, label, case_name = sampled_batch["image"], sampled_batch["label"], sampled_batch['case_name'][0]
        metric_i = val_single_volume(image, label, model, classes=args.num_classes, patch_size=[args.img_size, args.img_size],
                                      case=case_name, z_spacing=args.z_spacing)
        metric_list += np.array(metric_i)
    metric_list = metric_list / len(db_test)
    performance = np.mean(metric_list, axis=0)
    logging.info('Testing performance in val model: mean_dice : %f, best_dice : %f' % (performance, best_performance))
    return performance

def trainer_synapse(args, model, snapshot_path):
    logging.basicConfig(filename=snapshot_path + "/log.txt", level=logging.INFO,