Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
test_tta.py 7.35 KB
Newer Older
luopl's avatar
Initial commit  
luopl committed
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 
# Copyright (c) OpenMMLab. All rights reserved.
import os.path as osp
from unittest import TestCase

import mmcv
import pytest

from mmdet.datasets.transforms import *  # noqa
from mmdet.registry import TRANSFORMS


class TestMuitiScaleFlipAug(TestCase):

    def test_exception(self):
        with pytest.raises(TypeError):
            tta_transform = dict(
                type='TestTimeAug',
                transforms=[dict(type='Resize', keep_ratio=False)],
            )
            TRANSFORMS.build(tta_transform)

    def test_multi_scale_flip_aug(self):
        tta_transform = dict(
            type='TestTimeAug',
            transforms=[[
                dict(type='Resize', scale=scale, keep_ratio=False)
                for scale in [(256, 256), (512, 512), (1024, 1024)]
            ],
                        [
                            dict(
                                type='mmdet.PackDetInputs',
                                meta_keys=('img_id', 'img_path', 'ori_shape',
                                           'img_shape', 'scale_factor'))
                        ]])
        tta_module = TRANSFORMS.build(tta_transform)

        results = dict()
        img = mmcv.imread(
            osp.join(osp.dirname(__file__), '../data/color.jpg'), 'color')
        results['img_id'] = '1'
        results['img_path'] = 'data/color.jpg'
        results['img'] = img
        results['ori_shape'] = img.shape
        results['ori_height'] = img.shape[0]
        results['ori_width'] = img.shape[1]
        # Set initial values for default meta_keys
        results['pad_shape'] = img.shape
        results['scale_factor'] = 1.0

        tta_results = tta_module(results.copy())
        assert [img.shape
                for img in tta_results['inputs']] == [(3, 256, 256),
                                                      (3, 512, 512),
                                                      (3, 1024, 1024)]

        tta_transform = dict(
            type='TestTimeAug',
            transforms=[
                [
                    dict(type='Resize', scale=scale, keep_ratio=False)
                    for scale in [(256, 256), (512, 512), (1024, 1024)]
                ],
                [
                    dict(type='RandomFlip', prob=0., direction='horizontal'),
                    dict(type='RandomFlip', prob=1., direction='horizontal')
                ],
                [
                    dict(
                        type='mmdet.PackDetInputs',
                        meta_keys=('img_id', 'img_path', 'ori_shape',
                                   'img_shape', 'scale_factor', 'flip',
                                   'flip_direction'))
                ]
            ])
        tta_module = TRANSFORMS.build(tta_transform)
        tta_results: dict = tta_module(results.copy())
        assert [img.shape
                for img in tta_results['inputs']] == [(3, 256, 256),
                                                      (3, 256, 256),
                                                      (3, 512, 512),
                                                      (3, 512, 512),
                                                      (3, 1024, 1024),
                                                      (3, 1024, 1024)]
        assert [
            data_sample.metainfo['flip']
            for data_sample in tta_results['data_samples']
        ] == [False, True, False, True, False, True]

        tta_transform = dict(
            type='TestTimeAug',
            transforms=[[
                dict(type='Resize', scale=(512, 512), keep_ratio=False)
            ],
                        [
                            dict(
                                type='mmdet.PackDetInputs',
                                meta_keys=('img_id', 'img_path', 'ori_shape',
                                           'img_shape', 'scale_factor'))
                        ]])
        tta_module = TRANSFORMS.build(tta_transform)
        tta_results = tta_module(results.copy())
        assert [tta_results['inputs'][0].shape] == [(3, 512, 512)]

        tta_transform = dict(
            type='TestTimeAug',
            transforms=[
                [dict(type='Resize', scale=(512, 512), keep_ratio=False)],
                [
                    dict(type='RandomFlip', prob=0., direction='horizontal'),
                    dict(type='RandomFlip', prob=1., direction='horizontal')
                ],
                [
                    dict(
                        type='mmdet.PackDetInputs',
                        meta_keys=('img_id', 'img_path', 'ori_shape',
                                   'img_shape', 'scale_factor', 'flip',
                                   'flip_direction'))
                ]
            ])
        tta_module = TRANSFORMS.build(tta_transform)
        tta_results = tta_module(results.copy())
        assert [img.shape for img in tta_results['inputs']] == [(3, 512, 512),
                                                                (3, 512, 512)]
        assert [
            data_sample.metainfo['flip']
            for data_sample in tta_results['data_samples']
        ] == [False, True]

        tta_transform = dict(
            type='TestTimeAug',
            transforms=[[
                dict(type='Resize', scale_factor=r, keep_ratio=False)
                for r in [0.5, 1.0, 2.0]
            ],
                        [
                            dict(
                                type='mmdet.PackDetInputs',
                                meta_keys=('img_id', 'img_path', 'ori_shape',
                                           'img_shape', 'scale_factor'))
                        ]])
        tta_module = TRANSFORMS.build(tta_transform)
        tta_results = tta_module(results.copy())
        assert [img.shape for img in tta_results['inputs']] == [(3, 144, 256),
                                                                (3, 288, 512),
                                                                (3, 576, 1024)]

        tta_transform = dict(
            type='TestTimeAug',
            transforms=[
                [
                    dict(type='Resize', scale_factor=r, keep_ratio=True)
                    for r in [0.5, 1.0, 2.0]
                ],
                [
                    dict(type='RandomFlip', prob=0., direction='horizontal'),
                    dict(type='RandomFlip', prob=1., direction='horizontal')
                ],
                [
                    dict(
                        type='mmdet.PackDetInputs',
                        meta_keys=('img_id', 'img_path', 'ori_shape',
                                   'img_shape', 'scale_factor', 'flip',
                                   'flip_direction'))
                ]
            ])
        tta_module = TRANSFORMS.build(tta_transform)
        tta_results = tta_module(results.copy())
        assert [img.shape for img in tta_results['inputs']] == [(3, 144, 256),
                                                                (3, 144, 256),
                                                                (3, 288, 512),
                                                                (3, 288, 512),
                                                                (3, 576, 1024),
                                                                (3, 576, 1024)]
        assert [
            data_sample.metainfo['flip']
            for data_sample in tta_results['data_samples']
        ] == [False, True, False, True, False, True]