mirror of
https://github.com/hanzoai/anomalib.git
synced 2026-10-30 18:48:11 +00:00
zeekay
dda5b2a367
The previous commit was meant to remove .github/workflows and removed every tracked path instead — 892 files, the whole repo. This puts back the 874 that were never in scope; the eighteen workflow files stay gone, for the reasons that commit gives. Content is byte-identical to 9d2aec9 outside .github/workflows.
73 lines
2.8 KiB
Python
73 lines
2.8 KiB
Python
"""Unit Tests - Base Video Datamodules."""
|
|
|
|
# Copyright (C) 2023-2024 Intel Corporation
|
|
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
from dataclasses import fields
|
|
|
|
import pytest
|
|
import torch
|
|
|
|
from anomalib.data import AnomalibDataModule
|
|
from tests.unit.data.datamodule.base.base import _TestAnomalibDataModule
|
|
|
|
|
|
class _TestAnomalibVideoDatamodule(_TestAnomalibDataModule):
|
|
@staticmethod
|
|
@pytest.mark.parametrize("subset", ["train", "val", "test"])
|
|
def test_get_item_returns_correct_keys_and_shapes(datamodule: AnomalibDataModule, subset: str) -> None:
|
|
"""Test that the datamodule __getitem__ returns image, mask, label and boxes."""
|
|
# Get the dataloader.
|
|
dataloader = getattr(datamodule, f"{subset}_dataloader")()
|
|
|
|
# Get the first batch.
|
|
batch = next(iter(dataloader))
|
|
|
|
# Check that the batch has the correct keys.
|
|
expected_train_fields = {"image", "video_path", "frames", "last_frame", "original_image"}
|
|
expected_eval_fields = expected_train_fields | {"gt_label", "gt_mask"}
|
|
|
|
expected_fields = expected_train_fields if subset == "train" else expected_eval_fields
|
|
|
|
batch_fields = {field.name for field in fields(batch) if getattr(batch, field.name) is not None}
|
|
assert batch_fields == expected_fields
|
|
|
|
# Check that the batch has the correct shape.
|
|
assert batch.image.shape == (4, 2, 3, 256, 256)
|
|
assert len(batch.video_path) == 4
|
|
assert len(batch.frames) == 4
|
|
assert len(batch.last_frame) == 4
|
|
# We don't know the shape of the original image, so we only check that it is a list of 4 images.
|
|
assert batch.original_image.shape[0] == 4
|
|
|
|
if subset in {"val", "test"}:
|
|
assert len(batch.gt_label) == 4
|
|
assert batch.gt_mask.shape == (4, 256, 256)
|
|
assert batch.gt_mask.shape == (4, 256, 256)
|
|
|
|
@staticmethod
|
|
@pytest.mark.parametrize("subset", ["train", "val", "test"])
|
|
def test_item_dtype(subset: str, datamodule: AnomalibDataModule) -> None:
|
|
"""Test that the input tensor is of float type and scaled between 0-1."""
|
|
# Get the dataloader.
|
|
dataloader = getattr(datamodule, f"{subset}_dataloader")()
|
|
|
|
# Get the first batch.
|
|
batch = next(iter(dataloader))
|
|
clip = batch.image
|
|
assert clip.dtype == torch.float32
|
|
assert clip.min() >= 0
|
|
assert clip.max() <= 1
|
|
|
|
@staticmethod
|
|
@pytest.mark.parametrize("clip_length_in_frames", [1])
|
|
def test_single_frame_squeezed(datamodule: AnomalibDataModule) -> None:
|
|
"""Test that the temporal dimension is squeezed when the clip lenght is 1."""
|
|
# Get the dataloader.
|
|
dataloader = datamodule.train_dataloader()
|
|
|
|
# Get the first batch.
|
|
batch = next(iter(dataloader))
|
|
clip = batch.image
|
|
assert clip.shape == (4, 3, 256, 256)
|