Function evaluate_gen

test_generation.py:405–438

(opt, ref_pcs, logger)

def evaluate_gen(opt, ref_pcs, logger):

    if ref_pcs is None:
        _, test_dataset = get_dataset(opt.dataroot, opt.npoints, opt.category, use_mask=False)
        test_dataloader = torch.utils.data.DataLoader(test_dataset, batch_size=opt.batch_size,
                                                      shuffle=False, num_workers=int(opt.workers), drop_last=False)
        ref = []
        for data in tqdm(test_dataloader, total=len(test_dataloader), desc='Generating Samples'):
            x = data['test_points']
            m, s = data['mean'].float(), data['std'].float()

            ref.append(x*s + m)

        ref_pcs = torch.cat(ref, dim=0).contiguous()

    logger.info("Loading sample path: %s"
      % (opt.eval_path))
    sample_pcs = torch.load(opt.eval_path).contiguous()

    logger.info("Generation sample size:%s reference size: %s"
          % (sample_pcs.size(), ref_pcs.size()))


    # Compute metrics
    results = compute_all_metrics(sample_pcs, ref_pcs, opt.batch_size)
    results = {k: (v.cpu().detach().item()
                   if not isinstance(v, float) else v) for k, v in results.items()}

    pprint(results)
    logger.info(results)

    jsd = JSD(sample_pcs.numpy(), ref_pcs.numpy())
    pprint('JSD: {}'.format(jsd))
    logger.info('JSD: {}'.format(jsd))