File size: 14,465 Bytes
5f89bea |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 |
# coding=utf-8
# Copyright 2021 HuggingFace Inc.
#
# 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 json
import pathlib
import unittest
import numpy as np
from transformers.file_utils import is_torch_available, is_vision_available
from transformers.testing_utils import require_torch, require_vision, slow
from .test_feature_extraction_common import FeatureExtractionSavingTestMixin, prepare_image_inputs
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
from transformers import DetrFeatureExtractor
class DetrFeatureExtractionTester(unittest.TestCase):
def __init__(
self,
parent,
batch_size=7,
num_channels=3,
min_resolution=30,
max_resolution=400,
do_resize=True,
size=18,
max_size=1333, # by setting max_size > max_resolution we're effectively not testing this :p
do_normalize=True,
image_mean=[0.5, 0.5, 0.5],
image_std=[0.5, 0.5, 0.5],
):
self.parent = parent
self.batch_size = batch_size
self.num_channels = num_channels
self.min_resolution = min_resolution
self.max_resolution = max_resolution
self.do_resize = do_resize
self.size = size
self.max_size = max_size
self.do_normalize = do_normalize
self.image_mean = image_mean
self.image_std = image_std
def prepare_feat_extract_dict(self):
return {
"do_resize": self.do_resize,
"size": self.size,
"max_size": self.max_size,
"do_normalize": self.do_normalize,
"image_mean": self.image_mean,
"image_std": self.image_std,
}
def get_expected_values(self, image_inputs, batched=False):
"""
This function computes the expected height and width when providing images to DetrFeatureExtractor,
assuming do_resize is set to True with a scalar size.
"""
if not batched:
image = image_inputs[0]
if isinstance(image, Image.Image):
w, h = image.size
else:
h, w = image.shape[1], image.shape[2]
if w < h:
expected_height = int(self.size * h / w)
expected_width = self.size
elif w > h:
expected_height = self.size
expected_width = int(self.size * w / h)
else:
expected_height = self.size
expected_width = self.size
else:
expected_values = []
for image in image_inputs:
expected_height, expected_width = self.get_expected_values([image])
expected_values.append((expected_height, expected_width))
expected_height = max(expected_values, key=lambda item: item[0])[0]
expected_width = max(expected_values, key=lambda item: item[1])[1]
return expected_height, expected_width
@require_torch
@require_vision
class DetrFeatureExtractionTest(FeatureExtractionSavingTestMixin, unittest.TestCase):
feature_extraction_class = DetrFeatureExtractor if is_vision_available() else None
def setUp(self):
self.feature_extract_tester = DetrFeatureExtractionTester(self)
@property
def feat_extract_dict(self):
return self.feature_extract_tester.prepare_feat_extract_dict()
def test_feat_extract_properties(self):
feature_extractor = self.feature_extraction_class(**self.feat_extract_dict)
self.assertTrue(hasattr(feature_extractor, "image_mean"))
self.assertTrue(hasattr(feature_extractor, "image_std"))
self.assertTrue(hasattr(feature_extractor, "do_normalize"))
self.assertTrue(hasattr(feature_extractor, "do_resize"))
self.assertTrue(hasattr(feature_extractor, "size"))
self.assertTrue(hasattr(feature_extractor, "max_size"))
def test_batch_feature(self):
pass
def test_call_pil(self):
# Initialize feature_extractor
feature_extractor = self.feature_extraction_class(**self.feat_extract_dict)
# create random PIL images
image_inputs = prepare_image_inputs(self.feature_extract_tester, equal_resolution=False)
for image in image_inputs:
self.assertIsInstance(image, Image.Image)
# Test not batched input
encoded_images = feature_extractor(image_inputs[0], return_tensors="pt").pixel_values
expected_height, expected_width = self.feature_extract_tester.get_expected_values(image_inputs)
self.assertEqual(
encoded_images.shape,
(1, self.feature_extract_tester.num_channels, expected_height, expected_width),
)
# Test batched
expected_height, expected_width = self.feature_extract_tester.get_expected_values(image_inputs, batched=True)
encoded_images = feature_extractor(image_inputs, return_tensors="pt").pixel_values
self.assertEqual(
encoded_images.shape,
(
self.feature_extract_tester.batch_size,
self.feature_extract_tester.num_channels,
expected_height,
expected_width,
),
)
def test_call_numpy(self):
# Initialize feature_extractor
feature_extractor = self.feature_extraction_class(**self.feat_extract_dict)
# create random numpy tensors
image_inputs = prepare_image_inputs(self.feature_extract_tester, equal_resolution=False, numpify=True)
for image in image_inputs:
self.assertIsInstance(image, np.ndarray)
# Test not batched input
encoded_images = feature_extractor(image_inputs[0], return_tensors="pt").pixel_values
expected_height, expected_width = self.feature_extract_tester.get_expected_values(image_inputs)
self.assertEqual(
encoded_images.shape,
(1, self.feature_extract_tester.num_channels, expected_height, expected_width),
)
# Test batched
encoded_images = feature_extractor(image_inputs, return_tensors="pt").pixel_values
expected_height, expected_width = self.feature_extract_tester.get_expected_values(image_inputs, batched=True)
self.assertEqual(
encoded_images.shape,
(
self.feature_extract_tester.batch_size,
self.feature_extract_tester.num_channels,
expected_height,
expected_width,
),
)
def test_call_pytorch(self):
# Initialize feature_extractor
feature_extractor = self.feature_extraction_class(**self.feat_extract_dict)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.feature_extract_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test not batched input
encoded_images = feature_extractor(image_inputs[0], return_tensors="pt").pixel_values
expected_height, expected_width = self.feature_extract_tester.get_expected_values(image_inputs)
self.assertEqual(
encoded_images.shape,
(1, self.feature_extract_tester.num_channels, expected_height, expected_width),
)
# Test batched
encoded_images = feature_extractor(image_inputs, return_tensors="pt").pixel_values
expected_height, expected_width = self.feature_extract_tester.get_expected_values(image_inputs, batched=True)
self.assertEqual(
encoded_images.shape,
(
self.feature_extract_tester.batch_size,
self.feature_extract_tester.num_channels,
expected_height,
expected_width,
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize feature_extractors
feature_extractor_1 = self.feature_extraction_class(**self.feat_extract_dict)
feature_extractor_2 = self.feature_extraction_class(do_resize=False, do_normalize=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.feature_extract_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the feature extractor return the same tensors
encoded_images_with_method = feature_extractor_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = feature_extractor_2(image_inputs, return_tensors="pt")
assert torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
assert torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
@slow
def test_call_pytorch_with_coco_detection_annotations(self):
# prepare image and target
image = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png")
with open("./tests/fixtures/tests_samples/COCO/coco_annotations.txt", "r") as f:
target = json.loads(f.read())
target = {"image_id": 39769, "annotations": target}
# encode them
# TODO replace by facebook/detr-resnet-50
feature_extractor = DetrFeatureExtractor.from_pretrained("nielsr/detr-resnet-50")
encoding = feature_extractor(images=image, annotations=target, return_tensors="pt")
# verify pixel values
expected_shape = torch.Size([1, 3, 800, 1066])
self.assertEqual(encoding["pixel_values"].shape, expected_shape)
expected_slice = torch.tensor([0.2796, 0.3138, 0.3481])
assert torch.allclose(encoding["pixel_values"][0, 0, 0, :3], expected_slice, atol=1e-4)
# verify area
expected_area = torch.tensor([5887.9600, 11250.2061, 489353.8438, 837122.7500, 147967.5156, 165732.3438])
assert torch.allclose(encoding["target"][0]["area"], expected_area)
# verify boxes
expected_boxes_shape = torch.Size([6, 4])
self.assertEqual(encoding["target"][0]["boxes"].shape, expected_boxes_shape)
expected_boxes_slice = torch.tensor([0.5503, 0.2765, 0.0604, 0.2215])
assert torch.allclose(encoding["target"][0]["boxes"][0], expected_boxes_slice, atol=1e-3)
# verify image_id
expected_image_id = torch.tensor([39769])
assert torch.allclose(encoding["target"][0]["image_id"], expected_image_id)
# verify is_crowd
expected_is_crowd = torch.tensor([0, 0, 0, 0, 0, 0])
assert torch.allclose(encoding["target"][0]["iscrowd"], expected_is_crowd)
# verify class_labels
expected_class_labels = torch.tensor([75, 75, 63, 65, 17, 17])
assert torch.allclose(encoding["target"][0]["class_labels"], expected_class_labels)
# verify orig_size
expected_orig_size = torch.tensor([480, 640])
assert torch.allclose(encoding["target"][0]["orig_size"], expected_orig_size)
# verify size
expected_size = torch.tensor([800, 1066])
assert torch.allclose(encoding["target"][0]["size"], expected_size)
@slow
def test_call_pytorch_with_coco_panoptic_annotations(self):
# prepare image, target and masks_path
image = Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png")
with open("./tests/fixtures/tests_samples/COCO/coco_panoptic_annotations.txt", "r") as f:
target = json.loads(f.read())
target = {"file_name": "000000039769.png", "image_id": 39769, "segments_info": target}
masks_path = pathlib.Path("./tests/fixtures/tests_samples/COCO/coco_panoptic")
# encode them
# TODO replace by .from_pretrained facebook/detr-resnet-50-panoptic
feature_extractor = DetrFeatureExtractor(format="coco_panoptic")
encoding = feature_extractor(images=image, annotations=target, masks_path=masks_path, return_tensors="pt")
# verify pixel values
expected_shape = torch.Size([1, 3, 800, 1066])
self.assertEqual(encoding["pixel_values"].shape, expected_shape)
expected_slice = torch.tensor([0.2796, 0.3138, 0.3481])
assert torch.allclose(encoding["pixel_values"][0, 0, 0, :3], expected_slice, atol=1e-4)
# verify area
expected_area = torch.tensor([147979.6875, 165527.0469, 484638.5938, 11292.9375, 5879.6562, 7634.1147])
assert torch.allclose(encoding["target"][0]["area"], expected_area)
# verify boxes
expected_boxes_shape = torch.Size([6, 4])
self.assertEqual(encoding["target"][0]["boxes"].shape, expected_boxes_shape)
expected_boxes_slice = torch.tensor([0.2625, 0.5437, 0.4688, 0.8625])
assert torch.allclose(encoding["target"][0]["boxes"][0], expected_boxes_slice, atol=1e-3)
# verify image_id
expected_image_id = torch.tensor([39769])
assert torch.allclose(encoding["target"][0]["image_id"], expected_image_id)
# verify is_crowd
expected_is_crowd = torch.tensor([0, 0, 0, 0, 0, 0])
assert torch.allclose(encoding["target"][0]["iscrowd"], expected_is_crowd)
# verify class_labels
expected_class_labels = torch.tensor([17, 17, 63, 75, 75, 93])
assert torch.allclose(encoding["target"][0]["class_labels"], expected_class_labels)
# verify masks
expected_masks_sum = 822338
self.assertEqual(encoding["target"][0]["masks"].sum().item(), expected_masks_sum)
# verify orig_size
expected_orig_size = torch.tensor([480, 640])
assert torch.allclose(encoding["target"][0]["orig_size"], expected_orig_size)
# verify size
expected_size = torch.tensor([800, 1066])
assert torch.allclose(encoding["target"][0]["size"], expected_size)
|