YOLOv5 分类模型 预处理 OpenCV实现_yolov5怎么数据集预处理

YOLOv5 分类模型 预处理 OpenCV实现

flyfish

YOLOv5 分类模型 预处理 PIL 实现
YOLOv5 分类模型 OpenCV和PIL两者实现预处理的差异

YOLOv5 分类模型 数据集加载 1 样本处理
YOLOv5 分类模型 数据集加载 2 切片处理
YOLOv5 分类模型 数据集加载 3 自定义类别

YOLOv5 分类模型的预处理（1） Resize 和 CenterCrop
YOLOv5 分类模型的预处理（2）ToTensor 和 Normalize

YOLOv5 分类模型 Top 1和Top 5 指标说明
YOLOv5 分类模型 Top 1和Top 5 指标实现

判断图像是否是np.ndarray类型和维度

OpenCV读取一张图像时，类型类型就是<class 'numpy.ndarray'>，这里判断图像是否是np.ndarray类型
dim是dimension维度的缩写,shape属性的长度也是它的ndim
灰度图的shape为HW，二个维度
RGB图的shape为HWC，三个维度

def _is_numpy_image(img):
    return isinstance(img, np.ndarray) and (img.ndim in {2, 3})
1
2

实现ToTensor和Normalize

def totensor_normalize(img):
    print("preprocess:",img.shape)
    images = (img/255-mean)/std
    images = images.transpose((2, 0, 1))# HWC to CHW
    images = np.ascontiguousarray(images)
    return images
1
2
3
4
5
6

实现Resize

插值可以是以下参数

# 'nearest': cv2.INTER_NEAREST,
# 'bilinear': cv2.INTER_LINEAR,
# 'area': cv2.INTER_AREA,
# 'bicubic': cv2.INTER_CUBIC,
# 'lanczos': cv2.INTER_LANCZOS4
1
2
3
4
5

def resize(img, size, interpolation=cv2.INTER_LINEAR):
    r"""Resize the input numpy ndarray to the given size.
    Args:
        img (numpy ndarray): Image to be resized.
        size: like pytroch about size interpretation flyfish.
        interpolation (int, optional): Desired interpolation. Default is``cv2.INTER_LINEAR``  
    Returns:
        numpy Image: Resized image.like opencv
    """
    if not _is_numpy_image(img):
        raise TypeError('img should be numpy image. Got {}'.format(type(img)))
    if not (isinstance(size, int) or (isinstance(size, collections.abc.Iterable) and len(size) == 2)):
        raise TypeError('Got inappropriate size arg: {}'.format(size))
    h, w = img.shape[0], img.shape[1]

    if isinstance(size, int):
        if (w <= h and w == size) or (h <= w and h == size):
            return img
        if w < h:
            ow = size
            oh = int(size * h / w)
        else:
            oh = size
            ow = int(size * w / h)
    else:
        ow, oh = size[1], size[0]
    output = cv2.resize(img, dsize=(ow, oh), interpolation=interpolation)
    if img.shape[2] == 1:
        return output[:, :, np.newaxis]
    else:
        return output
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

实现CenterCrop

def crop(img, i, j, h, w):
    """Crop the given Image flyfish.
    Args:
        img (numpy ndarray): Image to be cropped.
        i: Upper pixel coordinate.
        j: Left pixel coordinate.
        h: Height of the cropped image.
        w: Width of the cropped image.
    Returns:
        numpy ndarray: Cropped image.
    """
    if not _is_numpy_image(img):
        raise TypeError('img should be numpy image. Got {}'.format(type(img)))

    return img[i:i + h, j:j + w, :]


def center_crop(img, output_size):
    if isinstance(output_size, numbers.Number):
        output_size = (int(output_size), int(output_size))
    h, w = img.shape[0:2]
    th, tw = output_size
    i = int(round((h - th) / 2.))
    j = int(round((w - tw) / 2.))
    return crop(img, i, j, th, tw)
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

完整

import time
from models.common import DetectMultiBackend
import os
import os.path
from typing import Any, Callable, cast, Dict, List, Optional, Tuple, Union
import cv2
import numpy as np
import collections
import torch
import numbers


classes_name=['n02086240', 'n02087394', 'n02088364', 'n02089973', 'n02093754', 'n02096294', 'n02099601', 'n02105641', 'n02111889', 'n02115641']

mean=[0.485, 0.456, 0.406]
std=[0.229, 0.224, 0.225]

def _is_numpy_image(img):
    return isinstance(img, np.ndarray) and (img.ndim in {2, 3})


def totensor_normalize(img):
    print("preprocess:",img.shape)
    images = (img/255-mean)/std
    images = images.transpose((2, 0, 1))# HWC to CHW
    images = np.ascontiguousarray(images)
    return images

def resize(img, size, interpolation=cv2.INTER_LINEAR):
    r"""Resize the input numpy ndarray to the given size.
    Args:
        img (numpy ndarray): Image to be resized.
        size: like pytroch about size interpretation flyfish.
        interpolation (int, optional): Desired interpolation. Default is``cv2.INTER_LINEAR``  
    Returns:
        numpy Image: Resized image.like opencv
    """
    if not _is_numpy_image(img):
        raise TypeError('img should be numpy image. Got {}'.format(type(img)))
    if not (isinstance(size, int) or (isinstance(size, collections.abc.Iterable) and len(size) == 2)):
        raise TypeError('Got inappropriate size arg: {}'.format(size))
    h, w = img.shape[0], img.shape[1]

    if isinstance(size, int):
        if (w <= h and w == size) or (h <= w and h == size):
            return img
        if w < h:
            ow = size
            oh = int(size * h / w)
        else:
            oh = size
            ow = int(size * w / h)
    else:
        ow, oh = size[1], size[0]
    output = cv2.resize(img, dsize=(ow, oh), interpolation=interpolation)
    if img.shape[2] == 1:
        return output[:, :, np.newaxis]
    else:
        return output

def crop(img, i, j, h, w):
    """Crop the given Image flyfish.
    Args:
        img (numpy ndarray): Image to be cropped.
        i: Upper pixel coordinate.
        j: Left pixel coordinate.
        h: Height of the cropped image.
        w: Width of the cropped image.
    Returns:
        numpy ndarray: Cropped image.
    """
    if not _is_numpy_image(img):
        raise TypeError('img should be numpy image. Got {}'.format(type(img)))

    return img[i:i + h, j:j + w, :]


def center_crop(img, output_size):
    if isinstance(output_size, numbers.Number):
        output_size = (int(output_size), int(output_size))
    h, w = img.shape[0:2]
    th, tw = output_size
    i = int(round((h - th) / 2.))
    j = int(round((w - tw) / 2.))
    return crop(img, i, j, th, tw)

class DatasetFolder:

    def __init__(
        self,
        root: str,

    ) -> None:
        self.root = root

        if classes_name is None or not classes_name:
            classes, class_to_idx = self.find_classes(self.root)
            print("not classes_name")

        else:
            classes = classes_name
            class_to_idx ={cls_name: i for i, cls_name in enumerate(classes)}
            print("is classes_name")

        print("classes:",classes)
        
        print("class_to_idx:",class_to_idx)
        samples = self.make_dataset(self.root, class_to_idx)

        self.classes = classes
        self.class_to_idx = class_to_idx
        self.samples = samples
        self.targets = [s[1] for s in samples]

    @staticmethod
    def make_dataset(
        directory: str,
        class_to_idx: Optional[Dict[str, int]] = None,

    ) -> List[Tuple[str, int]]:

        directory = os.path.expanduser(directory)

        if class_to_idx is None:
            _, class_to_idx = self.find_classes(directory)
        elif not class_to_idx:
            raise ValueError("'class_to_index' must have at least one entry to collect any samples.")

        instances = []
        available_classes = set()
        for target_class in sorted(class_to_idx.keys()):
            class_index = class_to_idx[target_class]
            target_dir = os.path.join(directory, target_class)
            if not os.path.isdir(target_dir):
                continue
            for root, _, fnames in sorted(os.walk(target_dir, followlinks=True)):
                for fname in sorted(fnames):
                    path = os.path.join(root, fname)
                    if 1:  # 验证:
                        item = path, class_index
                        instances.append(item)

                        if target_class not in available_classes:
                            available_classes.add(target_class)

        empty_classes = set(class_to_idx.keys()) - available_classes
        if empty_classes:
            msg = f"Found no valid file for the classes {', '.join(sorted(empty_classes))}. "

        return instances

    def find_classes(self, directory: str) -> Tuple[List[str], Dict[str, int]]:

        classes = sorted(entry.name for entry in os.scandir(directory) if entry.is_dir())
        if not classes:
            raise FileNotFoundError(f"Couldn't find any class folder in {directory}.")

        class_to_idx = {cls_name: i for i, cls_name in enumerate(classes)}
        return classes, class_to_idx

    def __getitem__(self, index: int) -> Tuple[Any, Any]:

        path, target = self.samples[index]
        sample = self.loader(path)

        return sample, target

    def __len__(self) -> int:
        return len(self.samples)

    def loader(self, path):
        print("path:", path)
        img = cv2.imread(path)  # BGR HWC
        img=cv2.cvtColor(img,cv2.COLOR_BGR2RGB)#RGB
        print("type:",type(img))
        return img


def time_sync():
    return time.time()



dataset = DatasetFolder(root="/media/flyfish/datasets/imagewoof/val")
weights = "/home/classes.pt"
device = "cpu"
model = DetectMultiBackend(weights, device=device, dnn=False, fp16=False)
model.eval()



def classify_transforms(img):
    img=resize(img,224)
    img=center_crop(img,224)
    img=totensor_normalize(img)
    return img;

pred, targets, loss, dt = [], [], 0, [0.0, 0.0, 0.0]
# current batch size =1
for i, (images, labels) in enumerate(dataset):
    print("i:", i)
    print(images.shape, labels)
    im = classify_transforms(images)


    images=torch.from_numpy(im).to(torch.float32) # numpy to tensor
    images = images.unsqueeze(0).to("cpu")
 
    print(images.shape)


        
    t1 = time_sync()
    images = images.to(device, non_blocking=True)
    t2 = time_sync()
    # dt[0] += t2 - t1

    y = model(images)
    y=y.numpy()
   
    print("y:", y)
    t3 = time_sync()
    # dt[1] += t3 - t2

    tmp1=y.argsort()[:,::-1][:, :5]
   
    print("tmp1:", tmp1)
    pred.append(tmp1)

    print("labels:", labels)

    
    targets.append(labels)

    print("for pred:", pred)  # list
    print("for targets:", targets)  # list

    # dt[2] += time_sync() - t3


pred, targets = np.concatenate(pred), np.array(targets)
print("pred:", pred)
print("pred:", pred.shape)
print("targets:", targets)
print("targets:", targets.shape)
correct = ((targets[:, None] == pred)).astype(np.float32)
print("correct:", correct.shape)
print("correct:", correct)
acc = np.stack((correct[:, 0], correct.max(1)), axis=1)  # (top1, top5) accuracy
print("acc:", acc.shape)
print("acc:", acc)
top = acc.mean(0)
print("top1:", top[0])
print("top5:", top[1])
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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254

结果

pred: [[0 3 6 2 1]
 [0 7 2 9 3]
 [0 5 6 2 9]
 ...
 [9 8 7 6 1]
 [9 3 6 7 0]
 [9 5 0 2 7]]
pred: (3929, 5)
targets: [0 0 0 ... 9 9 9]
targets: (3929,)
correct: (3929, 5)
correct: [[          1           0           0           0           0]
 [          1           0           0           0           0]
 [          1           0           0           0           0]
 ...
 [          1           0           0           0           0]
 [          1           0           0           0           0]
 [          1           0           0           0           0]]
acc: (3929, 2)
acc: [[          1           1]
 [          1           1]
 [          1           1]
 ...
 [          1           1]
 [          1           1]
 [          1           1]]
top1: 0.86230594
top5: 0.98167473
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