EasyCV/tests/datasets/detection/test_det_mix_dataset.py

134 lines
4.9 KiB
Python

# Copyright (c) Alibaba, Inc. and its affiliates.
import os
import time
import unittest
import torch
from tests.ut_config import (COCO_CLASSES, DET_DATA_SMALL_COCO_LOCAL,
IMG_NORM_CFG_255)
from easycv.datasets.builder import build_dataset
class DetImagesMixDatasetTest(unittest.TestCase):
def setUp(self):
print(('Testing %s.%s' % (type(self).__name__, self._testMethodName)))
def test_load_train(self):
img_scale = (640, 640)
scale_ratio = (0.1, 2)
train_pipeline = [
dict(type='MMMosaic', img_scale=img_scale, pad_val=114.0),
dict(
type='MMRandomAffine',
scaling_ratio_range=scale_ratio,
border=(-img_scale[0] // 2, -img_scale[1] // 2)),
dict(
type='MMMixUp', # s m x l; tiny nano will detele
img_scale=img_scale,
ratio_range=(0.8, 1.6),
pad_val=114.0),
dict(
type='MMPhotoMetricDistortion',
brightness_delta=32,
contrast_range=(0.5, 1.5),
saturation_range=(0.5, 1.5),
hue_delta=18),
dict(type='MMRandomFlip', flip_ratio=0.5),
dict(type='MMResize', keep_ratio=True),
dict(
type='MMPad',
pad_to_square=True,
pad_val=(114.0, 114.0, 114.0)),
dict(type='MMNormalize', **IMG_NORM_CFG_255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels'])
]
train_dataset = dict(
type='DetImagesMixDataset',
data_source=dict(
type='DetSourceCoco',
ann_file=os.path.join(DET_DATA_SMALL_COCO_LOCAL,
'instances_train2017_20.json'),
img_prefix=os.path.join(DET_DATA_SMALL_COCO_LOCAL,
'train2017'),
pipeline=[
dict(type='LoadImageFromFile', to_float32=True),
dict(type='LoadAnnotations', with_bbox=True)
],
classes=COCO_CLASSES,
filter_empty_gt=False,
iscrowd=False),
pipeline=train_pipeline,
dynamic_scale=img_scale)
dataset = build_dataset(train_dataset)
data_num = len(dataset)
s = time.time()
for data in dataset:
pass
t = time.time()
print(f'read data done {(t-s)/data_num}s per sample')
self.assertEqual(data['img'].data.shape, torch.Size([3, 640, 640]))
img_metas = data['img_metas'].data
self.assertIn('flip', img_metas)
self.assertIn('filename', img_metas)
self.assertIn('img_shape', img_metas)
self.assertIn('ori_img_shape', img_metas)
self.assertEqual(data['gt_bboxes'].shape, torch.Size([120, 4]))
self.assertEqual(data['gt_labels'].shape, torch.Size([120, 1]))
self.assertEqual(data_num, 20)
def test_load_test(self):
img_scale = (640, 640)
test_pipeline = [
dict(type='MMResize', img_scale=img_scale, keep_ratio=True),
dict(
type='MMPad',
pad_to_square=True,
pad_val=(114.0, 114.0, 114.0)),
dict(type='MMNormalize', **IMG_NORM_CFG_255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels'])
]
val_dataset = dict(
type='DetImagesMixDataset',
data_source=dict(
type='DetSourceCoco',
ann_file=os.path.join(DET_DATA_SMALL_COCO_LOCAL,
'instances_val2017_20.json'),
img_prefix=os.path.join(DET_DATA_SMALL_COCO_LOCAL, 'val2017'),
pipeline=[
dict(type='LoadImageFromFile', to_float32=True),
dict(type='LoadAnnotations', with_bbox=True)
],
classes=COCO_CLASSES,
filter_empty_gt=False,
iscrowd=True),
pipeline=test_pipeline,
dynamic_scale=None,
label_padding=False)
dataset = build_dataset(val_dataset)
data_num = len(dataset)
s = time.time()
for data in dataset:
pass
t = time.time()
print(f'read data done {(t-s)/data_num}s per sample')
self.assertEqual(data['img'].data.shape, torch.Size([3, 640, 640]))
img_metas = data['img_metas'].data
self.assertIn('filename', img_metas)
self.assertIn('img_shape', img_metas)
self.assertIn('ori_img_shape', img_metas)
self.assertEqual(data['gt_bboxes'].data.shape, torch.Size([16, 4]))
self.assertEqual(data['gt_labels'].data.shape, torch.Size([16]))
self.assertEqual(data_num, 20)
if __name__ == '__main__':
unittest.main()