1# Copyright 2019 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# ============================================================================== 15""" 16Testing the random vertical flip op in DE 17""" 18import numpy as np 19import mindspore.dataset as ds 20import mindspore.dataset.transforms.py_transforms 21import mindspore.dataset.transforms.c_transforms as ops 22import mindspore.dataset.vision.c_transforms as c_vision 23import mindspore.dataset.vision.py_transforms as py_vision 24from mindspore import log as logger 25from util import save_and_check_md5, visualize_list, visualize_image, diff_mse, \ 26 config_get_set_seed, config_get_set_num_parallel_workers 27 28GENERATE_GOLDEN = False 29 30DATA_DIR = ["../data/dataset/test_tf_file_3_images/train-0000-of-0001.data"] 31SCHEMA_DIR = "../data/dataset/test_tf_file_3_images/datasetSchema.json" 32 33 34def v_flip(image): 35 """ 36 Apply the random_vertical 37 """ 38 39 # with the seed provided in this test case, it will always flip. 40 # that's why we flip here too 41 image = image[::-1, :, :] 42 return image 43 44 45def test_random_vertical_op(plot=False): 46 """ 47 Test random_vertical with default probability 48 """ 49 logger.info("Test random_vertical") 50 51 # First dataset 52 data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 53 decode_op = c_vision.Decode() 54 random_vertical_op = c_vision.RandomVerticalFlip(1.0) 55 data1 = data1.map(operations=decode_op, input_columns=["image"]) 56 data1 = data1.map(operations=random_vertical_op, input_columns=["image"]) 57 58 # Second dataset 59 data2 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 60 data2 = data2.map(operations=decode_op, input_columns=["image"]) 61 62 num_iter = 0 63 for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1, output_numpy=True), 64 data2.create_dict_iterator(num_epochs=1, output_numpy=True)): 65 66 # with the seed value, we can only guarantee the first number generated 67 if num_iter > 0: 68 break 69 70 image_v_flipped = item1["image"] 71 image = item2["image"] 72 image_v_flipped_2 = v_flip(image) 73 74 mse = diff_mse(image_v_flipped, image_v_flipped_2) 75 assert mse == 0 76 logger.info("image_{}, mse: {}".format(num_iter + 1, mse)) 77 num_iter += 1 78 if plot: 79 visualize_image(image, image_v_flipped, mse, image_v_flipped_2) 80 81 82def test_random_vertical_valid_prob_c(): 83 """ 84 Test RandomVerticalFlip op with c_transforms: valid non-default input, expect to pass 85 """ 86 logger.info("test_random_vertical_valid_prob_c") 87 original_seed = config_get_set_seed(0) 88 original_num_parallel_workers = config_get_set_num_parallel_workers(1) 89 90 # Generate dataset 91 data = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 92 decode_op = c_vision.Decode() 93 random_horizontal_op = c_vision.RandomVerticalFlip(0.8) 94 data = data.map(operations=decode_op, input_columns=["image"]) 95 data = data.map(operations=random_horizontal_op, input_columns=["image"]) 96 97 filename = "random_vertical_01_c_result.npz" 98 save_and_check_md5(data, filename, generate_golden=GENERATE_GOLDEN) 99 100 # Restore config setting 101 ds.config.set_seed(original_seed) 102 ds.config.set_num_parallel_workers(original_num_parallel_workers) 103 104 105def test_random_vertical_valid_prob_py(): 106 """ 107 Test RandomVerticalFlip op with py_transforms: valid non-default input, expect to pass 108 """ 109 logger.info("test_random_vertical_valid_prob_py") 110 original_seed = config_get_set_seed(0) 111 original_num_parallel_workers = config_get_set_num_parallel_workers(1) 112 113 # Generate dataset 114 data = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 115 transforms = [ 116 py_vision.Decode(), 117 py_vision.RandomVerticalFlip(0.8), 118 py_vision.ToTensor() 119 ] 120 transform = mindspore.dataset.transforms.py_transforms.Compose(transforms) 121 data = data.map(operations=transform, input_columns=["image"]) 122 123 filename = "random_vertical_01_py_result.npz" 124 save_and_check_md5(data, filename, generate_golden=GENERATE_GOLDEN) 125 126 # Restore config setting 127 ds.config.set_seed(original_seed) 128 ds.config.set_num_parallel_workers(original_num_parallel_workers) 129 130 131def test_random_vertical_invalid_prob_c(): 132 """ 133 Test RandomVerticalFlip op in c_transforms: invalid input, expect to raise error 134 """ 135 logger.info("test_random_vertical_invalid_prob_c") 136 137 # Generate dataset 138 data = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 139 decode_op = c_vision.Decode() 140 try: 141 # Note: Valid range of prob should be [0.0, 1.0] 142 random_horizontal_op = c_vision.RandomVerticalFlip(1.5) 143 data = data.map(operations=decode_op, input_columns=["image"]) 144 data = data.map(operations=random_horizontal_op, input_columns=["image"]) 145 except ValueError as e: 146 logger.info("Got an exception in DE: {}".format(str(e))) 147 assert 'Input prob is not within the required interval of [0.0, 1.0].' in str(e) 148 149 150def test_random_vertical_invalid_prob_py(): 151 """ 152 Test RandomVerticalFlip op in py_transforms: invalid input, expect to raise error 153 """ 154 logger.info("test_random_vertical_invalid_prob_py") 155 156 # Generate dataset 157 data = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 158 try: 159 transforms = [ 160 py_vision.Decode(), 161 # Note: Valid range of prob should be [0.0, 1.0] 162 py_vision.RandomVerticalFlip(1.5), 163 py_vision.ToTensor() 164 ] 165 transform = mindspore.dataset.transforms.py_transforms.Compose(transforms) 166 data = data.map(operations=transform, input_columns=["image"]) 167 except ValueError as e: 168 logger.info("Got an exception in DE: {}".format(str(e))) 169 assert 'Input prob is not within the required interval of [0.0, 1.0].' in str(e) 170 171 172def test_random_vertical_comp(plot=False): 173 """ 174 Test test_random_vertical_flip and compare between python and c image augmentation ops 175 """ 176 logger.info("test_random_vertical_comp") 177 178 # First dataset 179 data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 180 decode_op = c_vision.Decode() 181 # Note: The image must be flipped if prob is set to be 1 182 random_horizontal_op = c_vision.RandomVerticalFlip(1) 183 data1 = data1.map(operations=decode_op, input_columns=["image"]) 184 data1 = data1.map(operations=random_horizontal_op, input_columns=["image"]) 185 186 # Second dataset 187 data2 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 188 transforms = [ 189 py_vision.Decode(), 190 # Note: The image must be flipped if prob is set to be 1 191 py_vision.RandomVerticalFlip(1), 192 py_vision.ToTensor() 193 ] 194 transform = mindspore.dataset.transforms.py_transforms.Compose(transforms) 195 data2 = data2.map(operations=transform, input_columns=["image"]) 196 197 images_list_c = [] 198 images_list_py = [] 199 for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1, output_numpy=True), 200 data2.create_dict_iterator(num_epochs=1, output_numpy=True)): 201 image_c = item1["image"] 202 image_py = (item2["image"].transpose(1, 2, 0) * 255).astype(np.uint8) 203 images_list_c.append(image_c) 204 images_list_py.append(image_py) 205 206 # Check if the output images are the same 207 mse = diff_mse(image_c, image_py) 208 assert mse < 0.001 209 if plot: 210 visualize_list(images_list_c, images_list_py, visualize_mode=2) 211 212def test_random_vertical_op_1(): 213 """ 214 Test RandomVerticalFlip with different fields. 215 """ 216 logger.info("Test RandomVerticalFlip with different fields.") 217 218 data = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False) 219 data = data.map(operations=ops.Duplicate(), input_columns=["image"], 220 output_columns=["image", "image_copy"], column_order=["image", "image_copy"]) 221 random_vertical_op = c_vision.RandomVerticalFlip(1.0) 222 decode_op = c_vision.Decode() 223 224 data = data.map(operations=decode_op, input_columns=["image"]) 225 data = data.map(operations=decode_op, input_columns=["image_copy"]) 226 data = data.map(operations=random_vertical_op, input_columns=["image", "image_copy"]) 227 228 num_iter = 0 229 for data1 in data.create_dict_iterator(num_epochs=1, output_numpy=True): 230 image = data1["image"] 231 image_copy = data1["image_copy"] 232 mse = diff_mse(image, image_copy) 233 logger.info("image_{}, mse: {}".format(num_iter + 1, mse)) 234 assert mse == 0 235 num_iter += 1 236 237 238if __name__ == "__main__": 239 test_random_vertical_op(plot=True) 240 test_random_vertical_valid_prob_c() 241 test_random_vertical_valid_prob_py() 242 test_random_vertical_invalid_prob_c() 243 test_random_vertical_invalid_prob_py() 244 test_random_vertical_comp(plot=True) 245 test_random_vertical_op_1() 246