transformers/tests/models/sam_hq/test_processor_samhq.py

# Copyright 2023 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import shutil
import tempfile
import unittest

import numpy as np

from transformers.testing_utils import require_torch, require_torchvision, require_vision
from transformers.utils import is_torch_available, is_vision_available

from ...test_processing_common import ProcessorTesterMixin, prepare_image_inputs


if is_vision_available():
    from PIL import Image

    from transformers import AutoProcessor, SamHQProcessor, SamImageProcessor

if is_torch_available():
    import torch


@require_vision
@require_torchvision
class SamHQProcessorTest(ProcessorTesterMixin, unittest.TestCase):
    processor_class = SamHQProcessor

    @classmethod
    def setUp(self):
        self.tmpdirname = tempfile.mkdtemp()
        image_processor = SamImageProcessor()
        processor = SamHQProcessor(image_processor)
        processor.save_pretrained(self.tmpdirname)

    def get_image_processor(self, **kwargs):
        return AutoProcessor.from_pretrained(self.tmpdirname, **kwargs).image_processor

    @classmethod
    def tearDown(self):
        shutil.rmtree(self.tmpdirname)

    # Processor tester class can't use ProcessorTesterMixin atm because the processor is atypical e.g. only contains an image processor
    def prepare_image_inputs(self):
        """This function prepares a list of PIL images."""
        return prepare_image_inputs()

    def prepare_mask_inputs(self):
        """This function prepares a list of PIL images, or a list of numpy arrays if one specifies numpify=True,
        or a list of PyTorch tensors if one specifies torchify=True.
        """
        mask_inputs = [np.random.randint(255, size=(30, 400), dtype=np.uint8)]
        mask_inputs = [Image.fromarray(x) for x in mask_inputs]
        return mask_inputs

    def test_tokenizer_defaults_preserved_by_kwargs(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_image_processor_defaults_preserved_by_image_kwargs(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_chat_template_save_loading(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_kwargs_overrides_default_image_processor_kwargs(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_kwargs_overrides_default_tokenizer_kwargs(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_unstructured_kwargs(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_unstructured_kwargs_batched(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_doubly_passed_kwargs(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_structured_kwargs_nested(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_structured_kwargs_nested_from_dict(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_save_load_pretrained_additional_features(self):
        self.skipTest("SamHQProcessor does not have a tokenizer")

    def test_image_processor_no_masks(self):
        image_processor = self.get_image_processor()

        processor = SamHQProcessor(image_processor=image_processor)

        image_input = self.prepare_image_inputs()

        input_feat_extract = image_processor(image_input, return_tensors="pt")
        input_processor = processor(images=image_input, return_tensors="pt")

        for key in input_feat_extract.keys():
            self.assertAlmostEqual(input_feat_extract[key].sum().item(), input_processor[key].sum().item(), delta=1e-2)

        for image in input_feat_extract.pixel_values:
            self.assertEqual(image.shape, (3, 1024, 1024))

        for original_size in input_feat_extract.original_sizes:
            np.testing.assert_array_equal(original_size, np.array([30, 400]))

        for reshaped_input_size in input_feat_extract.reshaped_input_sizes:
            np.testing.assert_array_equal(
                reshaped_input_size, np.array([77, 1024])
            )  # reshaped_input_size value is before padding

    def test_image_processor_with_masks(self):
        image_processor = self.get_image_processor()

        processor = SamHQProcessor(image_processor=image_processor)

        image_input = self.prepare_image_inputs()
        mask_input = self.prepare_mask_inputs()

        input_feat_extract = image_processor(images=image_input, segmentation_maps=mask_input, return_tensors="pt")
        input_processor = processor(images=image_input, segmentation_maps=mask_input, return_tensors="pt")

        for key in input_feat_extract.keys():
            self.assertAlmostEqual(input_feat_extract[key].sum().item(), input_processor[key].sum().item(), delta=1e-2)

        for label in input_feat_extract.labels:
            self.assertEqual(label.shape, (256, 256))

    @require_torch
    def test_post_process_masks(self):
        image_processor = self.get_image_processor()

        processor = SamHQProcessor(image_processor=image_processor)
        dummy_masks = [torch.ones((1, 3, 5, 5))]

        original_sizes = [[1764, 2646]]

        reshaped_input_size = [[683, 1024]]
        masks = processor.post_process_masks(dummy_masks, original_sizes, reshaped_input_size)
        self.assertEqual(masks[0].shape, (1, 3, 1764, 2646))

        masks = processor.post_process_masks(
            dummy_masks, torch.tensor(original_sizes), torch.tensor(reshaped_input_size)
        )
        self.assertEqual(masks[0].shape, (1, 3, 1764, 2646))

        # should also work with np
        dummy_masks = [np.ones((1, 3, 5, 5))]
        masks = processor.post_process_masks(dummy_masks, np.array(original_sizes), np.array(reshaped_input_size))

        self.assertEqual(masks[0].shape, (1, 3, 1764, 2646))

        dummy_masks = [[1, 0], [0, 1]]
        with self.assertRaises(ValueError):
            masks = processor.post_process_masks(dummy_masks, np.array(original_sizes), np.array(reshaped_input_size))