零基础入门CV实践教程v2 0 PDF

Datawhale
零基础⼊入⻔门CV实践教程
────────
# 版本⼁丨V1.0 #
By 阿⽔水、安晟、王程伟、张强、伊雪、王茂霖
Github Datawhalechina
http://www.datawhale.club
版本号版本时间修订内容
V1.0 2020.5.13
内容贡献者
Datawhale CV - Basline
Datawhale CV - Task01
成员介绍个⼈人主⻚页
Top https://www.zhihu.com/people/finlayliu
https://blog.csdn.net/u011583927
https://blog.csdn.net/weixin_40647819
Github https://github.com/QiangZiBro
Tianchi Github: https://github.com/mlw67
https://github.com/datawhalechina/team-learning/
1 Datawhale 零基础⼊入⻔门CV-Baseline讲解
1.1 Baseline思路路
1.2 运⾏行行环境及安装示例例
1.3 Baseline详解
2 Datawhale 零基础⼊入⻔门CV-Task1 赛题理理解
2.1 学习⽬目标
2.2 赛题数据
2.3 数据标签
2.4 评测指标
2.5 读取数据
2.6 解题思路路
2.7 本章⼩小节
3 Datawhale 零基础⼊入⻔门CV-Task2 数据读取与数据扩增
3.1 学习⽬目标
3.2 图像读取
3.2.1 Pillow
3.2.2 OpenCV
3.3 数据扩增⽅方法
3.3.1 数据扩增介绍
3.3.2 常⻅见的数据扩增⽅方法
3.3.3 常⽤用的数据扩增库
3.4 Pytorch读取数据
3.5 本章⼩小节
4 Datawhale 零基础⼊入⻔门CV-Task3 字符识别模型
4.1 学习⽬目标
4.2 CNN介绍
4.3 CNN发展
4.4 Pytorch构建CNN模型
4.5 本章⼩小节
5 Datawhale 零基础⼊入⻔门CV-Task4 模型训练与验证
5.1 学习⽬目标
5.2 构造验证集
5.3 模型训练与验证
5.4 模型保存与加载
5.5 模型调参流程
5.6 本章⼩小节
6 Datawhale 零基础⼊入⻔门CV-Task5 模型集成

6.1 学习⽬目标
6.2 集成学习⽅方法
6.3 深度学习中的集成学习
6.3.1 Dropout
6.3.2 TTA
6.3.3 Snapshot
6.4 结果后处理理
6.5 本章⼩小节
1 Datawhale 零基础⼊入⻔门CV-Baseline讲解
赛题：零基础⼊入⻔门CV赛事- 街景字符编码识别
https://tianchi.aliyun.com/competition/entrance/531795/information
1.1 Baseline思路路
CNN
1. Pytorch Dataset DataLoder
2. CNN Pytorch
3.
4.
1.2 运⾏行行环境及安装示例例
Python2/3 Pytorch1.x 4G GPU python3.7+ torch1.3.1gpu
1. Anaconda
$conda create -n py37_torch131 python=3.7
2. pytorch1.3.1
$source activate py37_torch131
$conda install pytorch=1.3.1 torchvision cudatoolkit=10.0
3.
$pip install jupyter tqdm opencv-python matplotlib pandas
4. notebook baseline
$jupyter-notebook
5. ../input/
1 import os, sys, glob, shutil, json
2 os.environ["CUDA_VISIBLE_DEVICES"] = '0'
3 import cv2
5 from PIL import Image

6 import numpy as np
8 from tqdm import tqdm, tqdm_notebook

9
10 import torch
11 torch.manual_seed(0)
12 torch.backends.cudnn.deterministic = False
13 torch.backends.cudnn.benchmark = True
14
15 import torchvision.models as models
16 import torchvision.transforms as transforms
17 import torchvision.datasets as datasets

18 import torch.nn as nn
19 import torch.nn.functional as F
20 import torch.optim as optim
21 from torch.autograd import Variable
22 from torch.utils.data.dataset import Dataset
1.3 Baseline详解
步骤1：定义好读取图像的Dataset
1 class SVHNDataset(Dataset):
2 def __init__(self, img_path, img_label, transform=None):
3 self.img_path = img_path
4 self.img_label = img_label
5 if transform is not None:
6 self.transform = transform
7 else:
8 self.transform = None
9
10 def __getitem__(self, index):
11 img = Image.open(self.img_path[index]).convert('RGB')
12
13 if self.transform is not None:
14 img = self.transform(img)
15
16 # 设置最⻓长的字符⻓长度为5个
17 lbl = np.array(self.img_label[index], dtype=np.int)

18 lbl = list(lbl) + (5 - len(lbl)) * [10]
19 return img, torch.from_numpy(np.array(lbl[:5]))

20
21 def __len__(self):
22 return len(self.img_path)
步骤2：定义好训练数据和验证数据的Dataset
1 train_path = glob.glob('../input/train/*.png')
2 train_path.sort()
3 train_json = json.load(open('../input/train.json'))
4 train_label = [train_json[x]['label'] for x in train_json]
5 print(len(train_path), len(train_label))
6
7 train_loader = torch.utils.data.DataLoader(
8 SVHNDataset(train_path, train_label,
9 transforms.Compose([
10 transforms.Resize((64, 128)),
11 transforms.RandomCrop((60, 120)),
12 transforms.ColorJitter(0.3, 0.3, 0.2),
13 transforms.RandomRotation(5),
14 transforms.ToTensor(),
15 transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
16 ])),
17 batch_size=40,
18 shuffle=True,
19 num_workers=10,
20 )
21
22 val_path = glob.glob('../input/val/*.png')
23 val_path.sort()
24 val_json = json.load(open('../input/val.json'))
25 val_label = [val_json[x]['label'] for x in val_json]

26 print(len(val_path), len(val_label))
27
28 val_loader = torch.utils.data.DataLoader(
29 SVHNDataset(val_path, val_label,
32 # transforms.ColorJitter(0.3, 0.3, 0.2),
33 # transforms.RandomRotation(5),
36 ])),
37 batch_size=40,
38 shuffle=False,
39 num_workers=10,
40 )
步骤3：定义好字符分类模型，使⽤用renset18的模型作为特征提取模块
1 class SVHN_Model1(nn.Module):
2 def __init__(self):
3 super(SVHN_Model1, self).__init__()
4
5 model_conv = models.resnet18(pretrained=True)
6 model_conv.avgpool = nn.AdaptiveAvgPool2d(1)
7 model_conv = nn.Sequential(*list(model_conv.children())[:-1])
8 self.cnn = model_conv
9
10 self.fc1 = nn.Linear(512, 11)


15
16 def forward(self, img):
17 feat = self.cnn(img)
18 # print(feat.shape)
19 feat = feat.view(feat.shape[0], -1)
20 c1 = self.fc1(feat)
25 return c1, c2, c3, c4, c5
步骤4：定义好训练、验证和预测模块
1 def train(train_loader, model, criterion, optimizer):

2 # 切换模型为训练模式
3 model.train()
4 train_loss = []
6 for i, (input, target) in enumerate(train_loader):

7 if use_cuda:
8 input = input.cuda()
9 target = target.cuda()
10
11 c0, c1, c2, c3, c4 = model(input)

12 loss = criterion(c0, target[:, 0]) + \
13 criterion(c1, target[:, 1]) + \

16 criterion(c4, target[:, 4])

17
18 # loss /= 6
19 optimizer.zero_grad()
20 loss.backward()
21 optimizer.step()
22
23 if i % 100 == 0:
24 print(loss.item())
25
26 train_loss.append(loss.item())
27 return np.mean(train_loss)
28
29 def validate(val_loader, model, criterion):
30 # 切换模型为预测模型
31 model.eval()
32 val_loss = []
33
34 # 不不记录模型梯度信息
35 with torch.no_grad():
36 for i, (input, target) in enumerate(val_loader):

37 if use_cuda:
39 target = target.cuda()
40
42 loss = criterion(c0, target[:, 0]) + \


46 criterion(c4, target[:, 4])
47 # loss /= 6
48 val_loss.append(loss.item())
49 return np.mean(val_loss)
50
51 def predict(test_loader, model, tta=10):
52 model.eval()
53 test_pred_tta = None
54
55 # TTA 次数
56 for _ in range(tta):
57 test_pred = []
58
60 for i, (input, target) in enumerate(test_loader):
61 if use_cuda:
63

65 output = np.concatenate([
66 c0.data.numpy(),
67 c1.data.numpy(),
68 c2.data.numpy(),
69 c3.data.numpy(),
70 c4.data.numpy()], axis=1)
71 test_pred.append(output)
72
73 test_pred = np.vstack(test_pred)
74 if test_pred_tta is None:
75 test_pred_tta = test_pred
76 else:
77 test_pred_tta += test_pred
78
79 return test_pred_tta
步骤5：迭代训练和验证模型
1 model = SVHN_Model1()
2 criterion = nn.CrossEntropyLoss()
3 optimizer = torch.optim.Adam(model.parameters(), 0.001)
4 best_loss = 1000.0
5
6 use_cuda = False
7 if use_cuda:
8 model = model.cuda()
9
10 for epoch in range(2):

11 train_loss = train(train_loader, model, criterion, optimizer, epoch)
12 val_loss = validate(val_loader, model, criterion)
13
14 val_label = [''.join(map(str, x)) for x in val_loader.dataset.img_label]
15 val_predict_label = predict(val_loader, model, 1)

16 val_predict_label = np.vstack([
17 val_predict_label[:, :11].argmax(1),
18 val_predict_label[:, 11:22].argmax(1),
22 ]).T
23 val_label_pred = []
24 for x in val_predict_label:
25 val_label_pred.append(''.join(map(str, x[x!=10])))
26
27 val_char_acc = np.mean(np.array(val_label_pred) == np.array(val_label))

28
29 print('Epoch: {0}, Train loss: {1} \t Val loss: {2}'.format(epoch, train_loss, val_loss))
30 print(val_char_acc)
31 # 记录下验证集精度
32 if val_loss < best_loss:
33 best_loss = val_loss
34 torch.save(model.state_dict(), './model.pt')
2 Epoch
Epoch: 0, Train loss: 3.1 Val loss: 3.4 0.3439
Epoch: 1, Train loss: 2.1 Val loss: 2.9 0.4346
步骤6：对测试集样本进⾏行行预测，⽣生成提交⽂文件
1 test_path = glob.glob('../input/test_a/*.png')
2 test_path.sort()
3 test_label = [[1]] * len(test_path)
4 print(len(val_path), len(val_label))
5
6 test_loader = torch.utils.data.DataLoader(
7 SVHNDataset(test_path, test_label,
10 transforms.RandomCrop((60, 120)),
11 # transforms.ColorJitter(0.3, 0.3, 0.2),
12 # transforms.RandomRotation(5),
15 ])),
16 batch_size=40,
17 shuffle=False,
18 num_workers=10,
19 )
20
21 test_predict_label = predict(test_loader, model, 1)
22
23 test_label = [''.join(map(str, x)) for x in test_loader.dataset.img_label]

24 test_predict_label = np.vstack([
25 test_predict_label[:, :11].argmax(1),
26 test_predict_label[:, 11:22].argmax(1),
30 ]).T
31
32 test_label_pred = []
33 for x in test_predict_label:
34 test_label_pred.append(''.join(map(str, x[x!=10])))
35
36 import pandas as pd
37 df_submit = pd.read_csv('../input/test_A_sample_submit.csv')
38 df_submit['file_code'] = test_label_pred
39 df_submit.to_csv('renset18.csv', index=None)
2 Epoch 0.33
关于Datawhale：Datawhale AI
Datawhale 计算机视觉
2 Datawhale 零基础⼊入⻔门CV-Task1 赛题理理解
赛题名称 CV
赛题⽬目标
赛题任务
SVHN paper
2.1 学习⽬目标
1.
2.
2.2 赛题数据
SVHN
: SVHN
Top
3W 1W
A 4W B 4W
2.3 数据标签
Field Description
top X
height
left Y
width
label
JSON
原始图⽚片图⽚片JSON标注
2.4 评测指标
< center> Score= / < /center>
2.5 读取数据
JSON
1 import json
3
4 <div STYLE="page-break-after: always;"></div>
5 # 数据标注处理理
6 def parse_json(d):
7 arr = np.array([
8 d['top'], d['height'], d['left'], d['width'], d['label']
9 ])
10 arr = arr.astype(int)
11 return arr
12
13 img = cv2.imread('../input/train/000000.png')
14 arr = parse_json(train_json['000000.png'])
15
16 plt.figure(figsize=(10, 10))
17 plt.subplot(1, arr.shape[1]+1, 1)
18 plt.imshow(img)
19 plt.xticks([]); plt.yticks([])
20
21 for idx in range(arr.shape[1]):
22 plt.subplot(1, arr.shape[1]+1, idx+2)

23 plt.imshow(img[arr[0, idx]:arr[0, idx]+arr[1, idx],arr[2, idx]:arr[2, idx]+arr[3, idx]])
24 plt.title(arr[4, idx])
25 plt.xticks([]); plt.yticks([])
2.6 解题思路路
2 3 4
字符属性图⽚片
42 2
241 3
7358 4
简单⼊入⻔门思路路：定⻓长字符识别
2-4 6
6 23 23XXXX 231 231XXX
6 11
专业字符识别思路路：不不定⻓长字符识别
CRNN
专业分类思路路：检测再识别
SSD
YOLO
2.7 本章⼩小节
3 Datawhale 零基础⼊入⻔门CV-Task2 数据读取与数据扩
增
Pytorch
3.1 学习⽬目标
1. Python Pytorch
2. Pytorch
3.2 图像读取
Python
Pillow OpenCV
3.2.1 Pillow
Pillow Python (PIL Pillow ipython
notebook
效果代码
from PIL import Image # Pillow # im =Image.open(cat.jpg')

from PIL import Image, ImageFilter im = Image.open('cat.jpg') # : im2 = i
m.filter(ImageFilter.BLUR) im2.save('blur.jpg', 'jpeg')
from PIL import Image # jpg : im = Image.open('c

at.jpg') im.thumbnail((w//2, h//2)) im.save('thumbnail.jpg', 'jpeg')
Pillow Pillow
Pillow https://pillow.readthedocs.io/en/stable/
3.2.2 OpenCV
OpenCV Intel OpenCV

OpenCV Pillow
效果代码
import cv2 # Opencv

img = cv2.imread('cat.jpg') # Opencv BRG, img = cv2.cvtColor(im
g, cv2.COLOR_BGR2RGB)
import cv2 # Opencv

img = cv2.imread('cat.jpg') img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) #
import cv2 # Opencv

img = cv2.imread('cat.jpg') img =cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) #
edges = cv2.Canny(img, 30, 70) cv2.imwrite('canny.jpg', edges)# Canny
OpenCV OpenCV OpenCV
OpenCV https://opencv.org/
OpenCV Github https://github.com/opencv/opencv
OpenCV https://github.com/opencv/opencv_contrib
3.3 数据扩增⽅方法
Pillow OpenCV
Data Augmentation
3.3.1 数据扩增介绍
数据扩增为什什么有⽤用？
A B
有哪些数据扩增⽅方法？
3.3.2 常⻅见的数据扩增⽅方法
torchvision
1. transforms.CenterCrop
2. transforms.ColorJitter
3. transforms.FiveCrop
4. transforms.Grayscale
5. transforms.Pad
6. transforms.RandomAffine
7. transforms.RandomCrop
8. transforms.RandomHorizontalFlip
9. transforms.RandomRotation
10. transforms.RandomVerticalFlip
3.3.3 常⽤用的数据扩增库
torchvision
https://github.com/pytorch/vision
pytorch torch
imgaug
https://github.com/aleju/imgaug
imgaug
albumentations
https://albumentations.readthedocs.io
3.4 Pytorch读取数据
Pytorch Pytorch
Pytorch Dataset DataLoder
2 import cv2
3
6
7 import torch

10
17 else:
19

22

25
26 # 原始SVHN中类别10为数字0
29

31
34
36 train_path.sort()
39
40 data = SVHNDataset(train_path, train_label,
42 # 缩放到固定尺⼨寸
44
45 # 随机颜⾊色变换
47
48 # 加⼊入随机旋转
50
51 # 将图⽚片转换为pytorch 的tesntor
52 # transforms.ToTensor(),
53
54 # 对图像像素进⾏行行归⼀一化
55 # transforms.Normalize([0.485,0.456,0.406],[0.229,0.224,0.225])
56 ]))
1 2 3
Dataset DataLoder Dataset DataLoder
1. Dataset
2. DataLoder Dataset
DataLoder
2 import cv2
3
6
7 import torch

10
17 else:
19

22

25
26 # 原始SVHN中类别10为数字0
29
31
34
36 train_path.sort()

39
41 SVHNDataset(train_path, train_label,


48 ])),
49 batch_size=10, # 每批样本个数
50 shuffle=False, # 是否打乱顺序
51 num_workers=10, # 读取的线程个数
52 )
53
54 for data in train_loader:
55 break
DataLoder Dataset data
torch.Size([10, 3, 64, 128]), torch.Size([10, 6])
batchsize * chanel * height * width
3.5 本章⼩小节
Pytorch
4 Datawhale 零基础⼊入⻔门CV-Task3 字符识别模型
Convolutional Neural Network, CNN
4.1 学习⽬目标
1. CNN
2. Pytorch CNN
4.2 CNN介绍
CNN CNN
CNN
CNN
CNN CNN convolution pooling
non-linear activation function fully connected layer

LeNet
5×5 stride=1
CNN
CNN
CNN End to End CNN
4.3 CNN发展
AlexNet VGG InceptionV3 ResNet
LeNet-5(1998)
AlexNet(2012)
VGG-16(2014)
Inception-v1 (2014)
ResNet-50 (2015)
4.4 Pytorch构建CNN模型
Pytorch CNN
Pytorch CNN Pytorch
CNN CNN 6
1 import torch
2 torch.manual_seed(0)
3 torch.backends.cudnn.deterministic = False
4 torch.backends.cudnn.benchmark = True
5
6 import torchvision.models as models
8 import torchvision.datasets as datasets

9 import torch.nn as nn
10 import torch.nn.functional as F
11 import torch.optim as optim
12 from torch.autograd import Variable
14
16 # 定义模型
20 # CNN提取特征模块
21 self.cnn = nn.Sequential(
22 nn.Conv2d(3, 16, kernel_size=(3, 3), stride=(2, 2)),
23 nn.ReLU(),
24 nn.MaxPool2d(2),
26 nn.ReLU(),
27 nn.MaxPool2d(2),
28 )
29 #
30 self.fc1 = nn.Linear(32*3*7, 11)

36

46 return c1, c2, c3, c4, c5, c6
47

2 # 损失函数
3 criterion = nn.CrossEntropyLoss()
5 # 优化器器
7
8 loss_plot, c0_plot = [], []
10 # 迭代10个Epoch
12 for data in train_loader:

13 c0, c1, c2, c3, c4, c5 = model(data[0])
14 loss = criterion(c0, data[1][:, 0]) + \
15 criterion(c1, data[1][:, 1]) + \

19 criterion(c5, data[1][:, 5])

20 loss /= 6
22 loss.backward()
23 optimizer.step()
24
25 loss_plot.append(loss.item())
26 c0_plot.append((c0.argmax(1) == data[1][:, 0]).sum().item()*1.0 / c0.shape[0])
27
28 print(epoch)
ImageNet
4
5 model_conv = models.resnet18(pretrained=True)
6 model_conv.avgpool = nn.AdaptiveAvgPool2d(1)
7 model_conv = nn.Sequential(*list(model_conv.children())[:-1])
8 self.cnn = model_conv
9
15
18 # print(feat.shape)
25 return c1, c2, c3, c4, c5
4.5 本章⼩小节
CNN CNN Pytorch CNN
5 Datawhale 零基础⼊入⻔门CV-Task4 模型训练与验证
CNN
1.
2.
3.
Pytorch
5.1 学习⽬目标
1.
2. Pytorch
5.2 构造验证集
Overfitting Underfitting
CNN
Model Complexity
“ ”
1. Train Set
2. Validation Set
3. Test Set
1. Hold-Out
2. Cross Validation CV
K K-1 1 K
K K
K CV K
3. BootStrap
在本次赛题中已经划分为验证集，因此选⼿手可以直接使⽤用训练集进⾏行行训练，并使⽤用验证集进⾏行行验证精度（当然你也
可以合并训练集和验证集，⾃自⾏行行划分验证集）。
-
-
“ ” - -
5.3 模型训练与验证
Pytorch CNN CNN
1.
2.
2 train_dataset,
3 batch_size=10,
4 shuffle=True,
5 num_workers=10,
6 )
7
8 val_loader = torch.utils.data.DataLoader(
9 val_dataset,
10 batch_size=10,
11 shuffle=False,
12 num_workers=10,
13 )
14
16 criterion = nn.CrossEntropyLoss (size_average=False)
18 best_loss = 1000.0

20 print('Epoch: ', epoch)
21
22 train(train_loader, model, criterion, optimizer, epoch)
23 val_loss = validate(val_loader, model, criterion)
24
25 # 记录下验证集精度
26 if val_loss < best_loss:
27 best_loss = val_loss
28 torch.save(model.state_dict(), './model.pt')
Epoch
1 def train(train_loader, model, criterion, optimizer, epoch):

2 # 切换模型为训练模式
3 model.train()
4
5 for i, (input, target) in enumerate(train_loader):

13 loss /= 6
15 loss.backward()
16 optimizer.step()
Epoch
1 def validate(val_loader, model, criterion):

2 # 切换模型为预测模型
3 model.eval()
4 val_loss = []
5
6 # 不不记录模型梯度信息
8 for i, (input, target) in enumerate(val_loader):


16 loss /= 6
17 val_loss.append(loss.item())
18 return np.mean(val_loss)
5.4 模型保存与加载
Pytorch
torch.save(model_object.state_dict(), 'model.pt')
model.load_state_dict(torch.load(' model.pt'))
5.5 模型调参流程
GPU
1. http://lamda.nju.edu.cn/weixs/project/CNNTricks/CNNTricks.html
2. http://karpathy.github.io/2019/04/25/recipe/
1. CNN
2. CNN
3.
5.6 本章⼩小节
关于Datawhale：Datawhale AI
Datawhale 计算机视觉
6 Datawhale 零基础⼊入⻔门CV-Task5 模型集成
6.1 学习⽬目标
1.
2.
6.2 集成学习⽅方法
Stacking Bagging Boosting
10 10 CNN
10 CNN
1.
2.
6.3 深度学习中的集成学习
6.3.1 Dropout
Dropout
Dropout CNN
Dropout

2 # 定义模型
6 # CNN提取特征模块
7 self.cnn = nn.Sequential(
9 nn.ReLU(),
10 nn.Dropout(0.25),
11 nn.MaxPool2d(2),
13 nn.ReLU(),
14 nn.Dropout(0.25),
15 nn.MaxPool2d(2),
16 )
17 #
24

34 return c1, c2, c3, c4, c5, c6
6.3.2 TTA
Test Time Augmentation TTA
1 2 3
1 def predict(test_loader, model, tta=10):
2 model.eval()
3 test_pred_tta = None
4 # TTA 次数
5 for _ in range(tta):
6 test_pred = []
7
9 for i, (input, target) in enumerate(test_loader):
11 output = np.concatenate([c0.data.numpy(), c1.data.numpy(),
12 c2.data.numpy(), c3.data.numpy(),
13 c4.data.numpy(), c5.data.numpy()], axis=1)
14 test_pred.append(output)
15
16 test_pred = np.vstack(test_pred)
17 if test_pred_tta is None:
18 test_pred_tta = test_pred
19 else:
20 test_pred_tta += test_pred
21
22 return test_pred_tta
6.3.3 Snapshot
10 CNN
CNN
Snapshot Ensembles cyclical learning rate
checkopint checkpoint
cyclical learning rate CNN

Snapshot
6.4 结果后处理理
1.
2.
6.5 本章⼩小节
1.
2. Dropout TTA
Datawhale
Datawhale AI
Datawhale “for the learner ”
AI “数据众智、众
创”
https://tianchi.aliyun.com/
L4
D UPS
10 ⽆无⼈人驾驶货运领域⾸首个独⻆角兽企业 2015
www.tusimple.com
机械⼯工业出版社华章公司——专注⾼高端IT图书出版25年年！ 1995
IT 25 30

零基础入门CV实践教程v2 0 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

零基础入门CV实践教程v2 0 PDF

Uploaded by

Copyright:

Available Formats

Datawhale

Tianchi Github: https://github.com/mlw67

6 Datawhale 零基础⼊入⻔门CV-Task5 模型集成

1. Pytorch Dataset DataLoder

$conda create -n py37_torch131 python=3.7

$source activate py37_torch131

$conda install pytorch=1.3.1 torchvision cudatoolkit=10.0

$pip install jupyter tqdm opencv-python matplotlib pandas

1 import os, sys, glob, shutil, json

5 from PIL import Image

8 from tqdm import tqdm, tqdm_notebook

16 import torchvision.transforms as transforms

17 import torchvision.datasets as datasets

21 from torch.autograd import Variable

22 from torch.utils.data.dataset import Dataset

2 def __init__(self, img_path, img_label, transform=None):

5 if transform is not None:

17 lbl = np.array(self.img_label[index], dtype=np.int)

19 return img, torch.from_numpy(np.array(lbl[:5]))

4 train_label = [train_json[x]['label'] for x in train_json]

25 val_label = [val_json[x]['label'] for x in val_json]

10 self.fc1 = nn.Linear(512, 11)

11 self.fc2 = nn.Linear(512, 11)

13 self.fc4 = nn.Linear(512, 11)

14 self.fc5 = nn.Linear(512, 11)

16 def forward(self, img):

19 feat = feat.view(feat.shape[0], -1)

25 return c1, c2, c3, c4, c5

1 def train(train_loader, model, criterion, optimizer):

6 for i, (input, target) in enumerate(train_loader):

11 c0, c1, c2, c3, c4 = model(input)

13 criterion(c1, target[:, 1]) + \

15 criterion(c3, target[:, 3]) + \

16 criterion(c4, target[:, 4])

36 for i, (input, target) in enumerate(val_loader):

41 c0, c1, c2, c3, c4 = model(input)

42 loss = criterion(c0, target[:, 0]) + \

44 criterion(c2, target[:, 2]) + \

45 criterion(c3, target[:, 3]) + \

60 for i, (input, target) in enumerate(test_loader):

64 c0, c1, c2, c3, c4 = model(input)

10 for epoch in range(2):

12 val_loss = validate(val_loader, model, criterion)

15 val_predict_label = predict(val_loader, model, 1)

27 val_char_acc = np.mean(np.array(val_label_pred) == np.array(val_label))

Epoch: 1, Train loss: 2.1 Val loss: 2.9 0.4346

23 test_label = [''.join(map(str, x)) for x in test_loader.dataset.img_label]

< center> Score= / < /center>

22 plt.subplot(1, arr.shape[1]+1, idx+2)

6 23 23XXXX 231 231XXX

Pillow Python (PIL Pillow ipython

from PIL import Image # Pillow # im =Image.open(cat.jpg')

from PIL import Image # jpg : im = Image.open('c

OpenCV Intel OpenCV

import cv2 # Opencv

import cv2 # Opencv

import cv2 # Opencv

OpenCV OpenCV OpenCV

OpenCV Github https://github.com/opencv/opencv

Pytorch Dataset DataLoder

1 import os, sys, glob, shutil, json

4 from PIL import Image

8 from torch.utils.data.dataset import Dataset

2 def init(self, img_path, img_label, transform=None):

20 def getitem(self, index):

20 def getitem(self, index):

32 self.fc3 = nn.Linear(3237, 11)