Пример #1
0
def test_precision_recall_per_image():
    ds_path = BASE_PATH / 'hair_drier_toaster_bear.json'
    df = cocoeval.generate_predictions(ds_path, ds_path)

    map, mar = cocoeval.precision_recall_per_image(df, df['img_path'].loc[0])
    assert map > 0
    assert mar > 0
Пример #2
0
def test_mean_average_precision_and_recall_per_class():
    ds_path = BASE_PATH / 'hair_drier_toaster_bear.json'
    df = cocoeval.generate_predictions(ds_path, ds_path)
    class_idx_metrics = cocoeval.mean_average_precision_and_recall_per_class(df)
    class_idxs = df['true_class_id'].unique()
    class_idxs = class_idxs[class_idxs > 0]
    print(class_idx_metrics)
    assert len(class_idx_metrics) == len(class_idxs)
Пример #3
0
def test_mean_average_precision_and_recall_per_class_with_name_min_score():
    ds_path = BASE_PATH / 'hair_drier_toaster_bear.json'

    ds = CocoDataset(ds_path)
    idx_class_name = {
        idx: cat_meta['name'] for idx, cat_meta in ds.cats.items()
    }

    df = cocoeval.generate_predictions(ds_path, ds_path)
    class_idx_metrics = cocoeval.mean_average_precision_and_recall_per_class(
        df, idx_class_dict=idx_class_name, min_score=0.5)
    class_idxs = df['true_class_id'].unique()
    class_idxs = class_idxs[class_idxs > 0]
    print(class_idx_metrics)
    assert len(class_idx_metrics) == len(class_idxs)
Пример #4
0
def test_generate_predictions():
    ds_path = BASE_PATH / 'hair_drier_toaster_bear.json'
    df = cocoeval.generate_predictions(ds_path, ds_path)
    assert len(df) > 0
    assert df['IOU'].mean() == 1.
Пример #5
0
def test_mean_average_precision_and_recall():
    ds_path = BASE_PATH / 'hair_drier_toaster_bear.json'
    df = cocoeval.generate_predictions(ds_path, ds_path)
    map, mar = cocoeval.mean_average_precision_and_recall(df)
    assert map == 1.
    assert mar == 1.
Пример #6
0
def test_generate_predictions_one_class():
    ds_path = BASE_PATH / 'hair_drier_toaster_bear.json'
    df = cocoeval.generate_predictions(ds_path, ds_path, category_idxs=[1])
    assert len(df) > 0
    assert df['IOU'].mean() > 0.