1# Copyright 2021 Huawei Technologies Co., Ltd 2# 3# Licensed under the Apache License, Version 2.0 (the "License"); 4# you may not use this file except in compliance with the License. 5# You may obtain a copy of the License at 6# 7# http://www.apache.org/licenses/LICENSE-2.0 8# 9# Unless required by applicable law or agreed to in writing, software 10# distributed under the License is distributed on an "AS IS" BASIS, 11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12# See the License for the specific language governing permissions and 13# limitations under the License. 14# ============================================================================== 15import os 16 17import json 18import matplotlib.pyplot as plt 19import numpy as np 20import pytest 21 22import mindspore.dataset as ds 23import mindspore.dataset.vision.c_transforms as c_vision 24 25 26DATASET_DIR = "../data/dataset/testCityscapesData/cityscapes" 27DATASET_DIR_TASK_JSON = "../data/dataset/testCityscapesData/cityscapes/testTaskJson" 28 29 30def test_cityscapes_basic(plot=False): 31 """ 32 Validate CityscapesDataset basic read. 33 """ 34 task = "color" # instance semantic polygon color 35 quality_mode = "fine" # fine coarse 36 usage = "train" # quality_mode=fine 'train', 'test', 'val', 'all' else 'train', 'train_extra', 'val', 'all' 37 data = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, 38 decode=True, shuffle=False) 39 count = 0 40 images_list = [] 41 task_list = [] 42 for item in data.create_dict_iterator(num_epochs=1, output_numpy=True): 43 images_list.append(item['image']) 44 task_list.append(item['task']) 45 count = count + 1 46 assert count == 5 47 if plot: 48 visualize_dataset(images_list, task_list, task) 49 50 51def visualize_dataset(images, labels, task): 52 """ 53 Helper function to visualize the dataset samples. 54 """ 55 if task == "polygon": 56 return 57 image_num = len(images) 58 for i in range(image_num): 59 plt.subplot(121) 60 plt.imshow(images[i]) 61 plt.title('Original') 62 plt.subplot(122) 63 plt.imshow(labels[i]) 64 plt.title(task) 65 plt.savefig('./cityscapes_{}_{}.jpg'.format(task, str(i))) 66 67 68def test_cityscapes_polygon(): 69 """ 70 Validate CityscapesDataset with task of polygon. 71 """ 72 usage = "train" 73 quality_mode = "fine" 74 task = "polygon" 75 data = ds.CityscapesDataset(DATASET_DIR_TASK_JSON, usage=usage, quality_mode=quality_mode, task=task) 76 count = 0 77 json_file = os.path.join(DATASET_DIR_TASK_JSON, "gtFine/train/aa/aa_000000_gtFine_polygons.json") 78 with open(json_file, "r") as f: 79 expected = json.load(f) 80 for item in data.create_dict_iterator(num_epochs=1, output_numpy=True): 81 task_dict = json.loads(str(item['task'], encoding="utf-8")) 82 assert task_dict == expected 83 count = count + 1 84 assert count == 1 85 86 87def test_cityscapes_basic_func(): 88 """ 89 Validate CityscapesDataset with repeat, batch and getter operation. 90 """ 91 # case 1: test num_samples 92 usage = "train" 93 quality_mode = "fine" 94 task = "color" 95 data1 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_samples=4) 96 num_iter1 = 0 97 for _ in data1.create_dict_iterator(num_epochs=1): 98 num_iter1 += 1 99 assert num_iter1 == 4 100 101 # case 2: test repeat 102 data2 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_samples=5) 103 data2 = data2.repeat(5) 104 num_iter2 = 0 105 for _ in data2.create_dict_iterator(num_epochs=1): 106 num_iter2 += 1 107 assert num_iter2 == 25 108 109 # case 3: test batch with drop_remainder=False 110 data3 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, decode=True) 111 resize_op = c_vision.Resize((100, 100)) 112 data3 = data3.map(operations=resize_op, input_columns=["image"], num_parallel_workers=1) 113 data3 = data3.map(operations=resize_op, input_columns=["task"], num_parallel_workers=1) 114 assert data3.get_dataset_size() == 5 115 assert data3.get_batch_size() == 1 116 data3 = data3.batch(batch_size=3) # drop_remainder is default to be False 117 assert data3.get_dataset_size() == 2 118 assert data3.get_batch_size() == 3 119 num_iter3 = 0 120 for _ in data3.create_dict_iterator(num_epochs=1): 121 num_iter3 += 1 122 assert num_iter3 == 2 123 124 # case 4: test batch with drop_remainder=True 125 data4 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, decode=True) 126 resize_op = c_vision.Resize((100, 100)) 127 data4 = data4.map(operations=resize_op, input_columns=["image"], num_parallel_workers=1) 128 data4 = data4.map(operations=resize_op, input_columns=["task"], num_parallel_workers=1) 129 assert data4.get_dataset_size() == 5 130 assert data4.get_batch_size() == 1 131 data4 = data4.batch(batch_size=3, drop_remainder=True) # the rest of incomplete batch will be dropped 132 assert data4.get_dataset_size() == 1 133 assert data4.get_batch_size() == 3 134 num_iter4 = 0 135 for _ in data4.create_dict_iterator(num_epochs=1): 136 num_iter4 += 1 137 assert num_iter4 == 1 138 139 # case 5: test get_col_names 140 data5 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, decode=True) 141 assert data5.get_col_names() == ["image", "task"] 142 143 144def test_cityscapes_sequential_sampler(): 145 """ 146 Test CityscapesDataset with SequentialSampler. 147 """ 148 task = "color" 149 quality_mode = "fine" 150 usage = "train" 151 152 num_samples = 5 153 sampler = ds.SequentialSampler(num_samples=num_samples) 154 data1 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, sampler=sampler) 155 data2 = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, 156 shuffle=False, num_samples=num_samples) 157 num_iter = 0 158 for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1, output_numpy=True), 159 data2.create_dict_iterator(num_epochs=1, output_numpy=True)): 160 np.testing.assert_array_equal(item1["task"], item2["task"]) 161 num_iter += 1 162 assert num_iter == num_samples 163 164 165def test_cityscapes_exception(): 166 """ 167 Validate CityscapesDataset with error parameters. 168 """ 169 task = "color" 170 quality_mode = "fine" 171 usage = "train" 172 173 error_msg_1 = "does not exist or is not a directory or permission denied!" 174 with pytest.raises(ValueError, match=error_msg_1): 175 ds.CityscapesDataset("NoExistsDir", usage=usage, quality_mode=quality_mode, task=task) 176 177 error_msg_2 = "sampler and shuffle cannot be specified at the same time" 178 with pytest.raises(RuntimeError, match=error_msg_2): 179 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, shuffle=False, 180 sampler=ds.PKSampler(3)) 181 182 error_msg_3 = "sampler and sharding cannot be specified at the same time" 183 with pytest.raises(RuntimeError, match=error_msg_3): 184 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_shards=2, 185 shard_id=0, sampler=ds.PKSampler(3)) 186 187 error_msg_4 = "num_shards is specified and currently requires shard_id as well" 188 with pytest.raises(RuntimeError, match=error_msg_4): 189 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_shards=10) 190 191 error_msg_5 = "shard_id is specified but num_shards is not" 192 with pytest.raises(RuntimeError, match=error_msg_5): 193 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, shard_id=0) 194 195 error_msg_6 = "Input shard_id is not within the required interval" 196 with pytest.raises(ValueError, match=error_msg_6): 197 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_shards=5, shard_id=-1) 198 with pytest.raises(ValueError, match=error_msg_6): 199 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_shards=5, shard_id=5) 200 with pytest.raises(ValueError, match=error_msg_6): 201 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_shards=2, shard_id=5) 202 203 error_msg_7 = "num_parallel_workers exceeds" 204 with pytest.raises(ValueError, match=error_msg_7): 205 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, shuffle=False, 206 num_parallel_workers=0) 207 with pytest.raises(ValueError, match=error_msg_7): 208 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, shuffle=False, 209 num_parallel_workers=256) 210 with pytest.raises(ValueError, match=error_msg_7): 211 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, shuffle=False, 212 num_parallel_workers=-2) 213 214 error_msg_8 = "Argument shard_id" 215 with pytest.raises(TypeError, match=error_msg_8): 216 ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task, num_shards=2, shard_id="0") 217 218 def exception_func(item): 219 raise Exception("Error occur!") 220 221 try: 222 data = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task) 223 data = data.map(operations=exception_func, input_columns=["image"], num_parallel_workers=1) 224 num_rows = 0 225 for _ in data.create_dict_iterator(): 226 num_rows += 1 227 assert False 228 except RuntimeError as e: 229 assert "map operation: [PyFunc] failed. The corresponding data files:" in str(e) 230 231 try: 232 data = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task) 233 data = data.map(operations=exception_func, input_columns=["image"], num_parallel_workers=1) 234 num_rows = 0 235 for _ in data.create_dict_iterator(): 236 num_rows += 1 237 assert False 238 except RuntimeError as e: 239 assert "map operation: [PyFunc] failed. The corresponding data files:" in str(e) 240 241 242def test_cityscapes_param(): 243 """ 244 Validate CityscapesDataset with basic parameters like usage, quality_mode and task. 245 """ 246 def test_config(usage="train", quality_mode="fine", task="color"): 247 try: 248 data = ds.CityscapesDataset(DATASET_DIR, usage=usage, quality_mode=quality_mode, task=task) 249 num_rows = 0 250 for _ in data.create_dict_iterator(num_epochs=1, output_numpy=True): 251 num_rows += 1 252 except (ValueError, TypeError, RuntimeError) as e: 253 return str(e) 254 return num_rows 255 256 assert test_config(usage="train") == 5 257 assert test_config(usage="test") == 1 258 assert test_config(usage="val") == 1 259 assert test_config(usage="all") == 7 260 assert "usage is not within the valid set of ['train', 'test', 'val', 'all']" \ 261 in test_config("invalid", "fine", "instance") 262 assert "Argument usage with value ['list'] is not of type [<class 'str'>]" \ 263 in test_config(["list"], "fine", "instance") 264 assert "quality_mode is not within the valid set of ['fine', 'coarse']" \ 265 in test_config("train", "invalid", "instance") 266 assert "Argument quality_mode with value ['list'] is not of type [<class 'str'>]" \ 267 in test_config("train", ["list"], "instance") 268 assert "task is not within the valid set of ['instance', 'semantic', 'polygon', 'color']." \ 269 in test_config("train", "fine", "invalid") 270 assert "Argument task with value ['list'] is not of type [<class 'str'>], but got <class 'list'>." \ 271 in test_config("train", "fine", ["list"]) 272 273 274if __name__ == "__main__": 275 test_cityscapes_basic() 276 test_cityscapes_polygon() 277 test_cityscapes_basic_func() 278 test_cityscapes_sequential_sampler() 279 test_cityscapes_exception() 280 test_cityscapes_param() 281