Function evaluation

text2motion/tools/evaluation.py:153–238

(log_file)

def evaluation(log_file):
    with open(log_file, 'w') as f:
        all_metrics = OrderedDict({'Matching Score': OrderedDict({}),
                                   'R_precision': OrderedDict({}),
                                   'FID': OrderedDict({}),
                                   'Diversity': OrderedDict({}),
                                   'MultiModality': OrderedDict({})})
        for replication in range(replication_times):
            motion_loaders = {}
            mm_motion_loaders = {}
            motion_loaders['ground truth'] = gt_loader
            for motion_loader_name, motion_loader_getter in eval_motion_loaders.items():
                motion_loader, mm_motion_loader = motion_loader_getter()
                motion_loaders[motion_loader_name] = motion_loader
                mm_motion_loaders[motion_loader_name] = mm_motion_loader

            print(f'==================== Replication {replication} ====================')
            print(f'==================== Replication {replication} ====================', file=f, flush=True)
            print(f'Time: {datetime.now()}')
            print(f'Time: {datetime.now()}', file=f, flush=True)
            mat_score_dict, R_precision_dict, acti_dict = evaluate_matching_score(motion_loaders, f)

            print(f'Time: {datetime.now()}')
            print(f'Time: {datetime.now()}', file=f, flush=True)
            fid_score_dict = evaluate_fid(gt_loader, acti_dict, f)

            print(f'Time: {datetime.now()}')
            print(f'Time: {datetime.now()}', file=f, flush=True)
            div_score_dict = evaluate_diversity(acti_dict, f)

            print(f'Time: {datetime.now()}')
            print(f'Time: {datetime.now()}', file=f, flush=True)
            mm_score_dict = evaluate_multimodality(mm_motion_loaders, f)

            print(f'!!! DONE !!!')
            print(f'!!! DONE !!!', file=f, flush=True)

            for key, item in mat_score_dict.items():
                if key not in all_metrics['Matching Score']:
                    all_metrics['Matching Score'][key] = [item]
                else:
                    all_metrics['Matching Score'][key] += [item]

            for key, item in R_precision_dict.items():
                if key not in all_metrics['R_precision']:
                    all_metrics['R_precision'][key] = [item]
                else:
                    all_metrics['R_precision'][key] += [item]

            for key, item in fid_score_dict.items():
                if key not in all_metrics['FID']:
                    all_metrics['FID'][key] = [item]
                else:
                    all_metrics['FID'][key] += [item]

            for key, item in div_score_dict.items():
                if key not in all_metrics['Diversity']:
                    all_metrics['Diversity'][key] = [item]