• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 CenterCrop op in DE
17"""
18import numpy as np
19import mindspore.dataset as ds
20import mindspore.dataset.transforms.py_transforms
21import mindspore.dataset.vision.c_transforms as vision
22import mindspore.dataset.vision.py_transforms as py_vision
23from mindspore import log as logger
24from util import diff_mse, visualize_list, save_and_check_md5
25
26GENERATE_GOLDEN = False
27
28DATA_DIR = ["../data/dataset/test_tf_file_3_images/train-0000-of-0001.data"]
29SCHEMA_DIR = "../data/dataset/test_tf_file_3_images/datasetSchema.json"
30
31
32def test_center_crop_op(height=375, width=375, plot=False):
33    """
34    Test CenterCrop
35    """
36    logger.info("Test CenterCrop")
37
38    # First dataset
39    data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"])
40    decode_op = vision.Decode()
41    # 3 images [375, 500] [600, 500] [512, 512]
42    center_crop_op = vision.CenterCrop([height, width])
43    data1 = data1.map(operations=decode_op, input_columns=["image"])
44    data1 = data1.map(operations=center_crop_op, input_columns=["image"])
45
46    # Second dataset
47    data2 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"])
48    data2 = data2.map(operations=decode_op, input_columns=["image"])
49
50    image_cropped = []
51    image = []
52    for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1, output_numpy=True),
53                            data2.create_dict_iterator(num_epochs=1, output_numpy=True)):
54        image_cropped.append(item1["image"].copy())
55        image.append(item2["image"].copy())
56    if plot:
57        visualize_list(image, image_cropped)
58
59
60def test_center_crop_md5(height=375, width=375):
61    """
62    Test CenterCrop
63    """
64    logger.info("Test CenterCrop")
65
66    # First dataset
67    data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False)
68    decode_op = vision.Decode()
69    # 3 images [375, 500] [600, 500] [512, 512]
70    center_crop_op = vision.CenterCrop([height, width])
71    data1 = data1.map(operations=decode_op, input_columns=["image"])
72    data1 = data1.map(operations=center_crop_op, input_columns=["image"])
73    # Compare with expected md5 from images
74    filename = "center_crop_01_result.npz"
75    save_and_check_md5(data1, filename, generate_golden=GENERATE_GOLDEN)
76
77
78def test_center_crop_comp(height=375, width=375, plot=False):
79    """
80    Test CenterCrop between python and c image augmentation
81    """
82    logger.info("Test CenterCrop")
83
84    # First dataset
85    data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False)
86    decode_op = vision.Decode()
87    center_crop_op = vision.CenterCrop([height, width])
88    data1 = data1.map(operations=decode_op, input_columns=["image"])
89    data1 = data1.map(operations=center_crop_op, input_columns=["image"])
90
91    # Second dataset
92    data2 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False)
93    transforms = [
94        py_vision.Decode(),
95        py_vision.CenterCrop([height, width]),
96        py_vision.ToTensor()
97    ]
98    transform = mindspore.dataset.transforms.py_transforms.Compose(transforms)
99    data2 = data2.map(operations=transform, input_columns=["image"])
100
101    image_c_cropped = []
102    image_py_cropped = []
103    for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1, output_numpy=True),
104                            data2.create_dict_iterator(num_epochs=1, output_numpy=True)):
105        c_image = item1["image"]
106        py_image = (item2["image"].transpose(1, 2, 0) * 255).astype(np.uint8)
107        # Note: The images aren't exactly the same due to rounding error
108        assert diff_mse(py_image, c_image) < 0.001
109        image_c_cropped.append(c_image.copy())
110        image_py_cropped.append(py_image.copy())
111    if plot:
112        visualize_list(image_c_cropped, image_py_cropped, visualize_mode=2)
113
114
115def test_crop_grayscale(height=375, width=375):
116    """
117    Test that centercrop works with pad and grayscale images
118    """
119
120    # Note: image.transpose performs channel swap to allow py transforms to
121    # work with c transforms
122    transforms = [
123        py_vision.Decode(),
124        py_vision.Grayscale(1),
125        py_vision.ToTensor(),
126        (lambda image: (image.transpose(1, 2, 0) * 255).astype(np.uint8))
127    ]
128
129    transform = mindspore.dataset.transforms.py_transforms.Compose(transforms)
130    data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False)
131    data1 = data1.map(operations=transform, input_columns=["image"])
132
133    # If input is grayscale, the output dimensions should be single channel
134    crop_gray = vision.CenterCrop([height, width])
135    data1 = data1.map(operations=crop_gray, input_columns=["image"])
136
137    for item1 in data1.create_dict_iterator(num_epochs=1, output_numpy=True):
138        c_image = item1["image"]
139
140        # Check that the image is grayscale
141        assert (c_image.ndim == 3 and c_image.shape[2] == 1)
142
143
144def test_center_crop_errors():
145    """
146    Test that CenterCropOp errors with bad input
147    """
148    try:
149        test_center_crop_op(16777216, 16777216)
150    except RuntimeError as e:
151        assert "CenterCropOp padding size is more than 3 times the original size" in \
152               str(e)
153
154
155if __name__ == "__main__":
156    test_center_crop_op(600, 600, plot=True)
157    test_center_crop_op(300, 600)
158    test_center_crop_op(600, 300)
159    test_center_crop_md5()
160    test_center_crop_comp(plot=True)
161    test_crop_grayscale()
162