赛题地址
前期环境
运行环境及安装
运行环境
-
python3.7
-
pytorch1.3.1
-
有GPU
首先在Anaconda中创建一个专门用于本次练习赛的虚拟环境。
$conda create -n pytorch_gpu python=3.7
激活环境,并安装pytorch1.3.1
$source activate pytorch_gpu
$conda install pytorch=1.3.1 torchvision cudatoolkit=10.0
一键安装所需其它依赖库
$pip install jupyter tqdm opencv-python matplotlib pandas
预训练
首先导入常用的包
import os, sys, glob, shutil, json
os.environ["CUDA_VISIBLE_DEVICES"] = '0'
import cv2
from PIL import Image
import numpy as np
from tqdm import tqdm, tqdm_notebook
import torch
torch.manual_seed(0)
torch.backends.cudnn.deterministic = False
torch.backends.cudnn.benchmark = True
import torchvision.models as models
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.autograd import Variable
from torch.utils.data.dataset import Dataset
步骤1:定义好读取图像的Dataset
class SVHNDataset(Dataset):
def __init__(self, img_path, img_label, transform=None):
self.img_path = img_path
self.img_label = img_label
if transform is not None:
self.transform = transform
else:
self.transform = None
def __getitem__(self, index):
img = Image.open(self.img_path[index]).convert('RGB')
if self.transform is not None:
img = self.transform(img)
# 设置最长的字符长度为5个
lbl = np.array(self.img_label[index], dtype=np.int)
lbl = list(lbl) + (5 - len(lbl)) * [10]
return img, torch.from_numpy(np.array(lbl[:5]))
def __len__(self):
return len(self.img_path)
步骤2:定义好训练数据和验证数据的Dataset
train_path = glob.glob('E:\python-project\deep-learning\cv-stree\mchar_val/*.png')
train_path.sort()
train_json = json.load(open('E:\python-project\deep-learning\cv-stree\train.json'))
train_label = [train_json[x]['label'] for x in train_json]
print(len(train_path), len(train_label))
train_loader = torch.utils.data.DataLoader(
SVHNDataset(train_path, train_label,
transforms.Compose([
transforms.Resize((64, 128)),
transforms.RandomCrop((60, 120)),
transforms.ColorJitter(0.3, 0.3, 0.2),
transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])),
batch_size=40,
shuffle=True,
num_workers=10,
)
val_path = glob.glob('E:\python-project\deep-learning\cv-stree\mchar_val/*.png')
val_path.sort()
val_json = json.load(open('E:\python-project\deep-learning\cv-stree\val.json'))
val_label = [val_json[x]['label'] for x in val_json]
print(len(val_path), len(val_label))
val_loader = torch.utils.data.DataLoader(
SVHNDataset(val_path, val_label,
transforms.Compose([
transforms.Resize((60, 120)),
# transforms.ColorJitter(0.3, 0.3, 0.2),
# transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])),
batch_size=40,
shuffle=False,
num_workers=10,
)
步骤3:定义好字符分类模型,使用renset18的模型作为特征提取模块
class SVHN_Model1(nn.Module):
def __init__(self):
super(SVHN_Model1, self).__init__()
model_conv = models.resnet18(pretrained=True)
model_conv.avgpool = nn.AdaptiveAvgPool2d(1)
model_conv = nn.Sequential(*list(model_conv.children())[:-1])
self.cnn = model_conv
self.fc1 = nn.Linear(512, 11)
self.fc2 = nn.Linear(512, 11)
self.fc3 = nn.Linear(512, 11)
self.fc4 = nn.Linear(512, 11)
self.fc5 = nn.Linear(512, 11)
def forward(self, img):
feat = self.cnn(img)
# print(feat.shape)
feat = feat.view(feat.shape[0], -1)
c1 = self.fc1(feat)
c2 = self.fc2(feat)
c3 = self.fc3(feat)
c4 = self.fc4(feat)
c5 = self.fc5(feat)
return c1, c2, c3, c4, c5
步骤4:定义好训练、验证和预测模块
def train(train_loader, model, criterion, optimizer):
# 切换模型为训练模式
model.train()
train_loss = []
for i, (input, target) in enumerate(train_loader):
if use_cuda:
input = input.cuda()
target = target.cuda()
c0, c1, c2, c3, c4 = model(input)
loss = criterion(c0, target[:, 0]) + \
criterion(c1, target[:, 1]) + \
criterion(c2, target[:, 2]) + \
criterion(c3, target[:, 3]) + \
criterion(c4, target[:, 4])
# loss /= 6
optimizer.zero_grad()
loss.backward()
optimizer.step()
if i % 100 == 0:
print(loss.item())
train_loss.append(loss.item())
return np.mean(train_loss)
def validate(val_loader, model, criterion):
# 切换模型为预测模型
model.eval()
val_loss = []
# 不记录模型梯度信息
with torch.no_grad():
for i, (input, target) in enumerate(val_loader):
if use_cuda:
input = input.cuda()
target = target.cuda()
c0, c1, c2, c3, c4 = model(input)
loss = criterion(c0, target[:, 0]) + \
criterion(c1, target[:, 1]) + \
criterion(c2, target[:, 2]) + \
criterion(c3, target[:, 3]) + \
criterion(c4, target[:, 4])
# loss /= 6
val_loss.append(loss.item())
return np.mean(val_loss)
def predict(test_loader, model, tta=10):
model.eval()
test_pred_tta = None
# TTA 次数
for _ in range(tta):
test_pred = []
with torch.no_grad():
for i, (input, target) in enumerate(test_loader):
if use_cuda:
input = input.cuda()
c0, c1, c2, c3, c4 = model(input)
output = np.concatenate([
c0.data.numpy(),
c1.data.numpy(),
c2.data.numpy(),
c3.data.numpy(),
c4.data.numpy()], axis=1)
test_pred.append(output)
test_pred = np.vstack(test_pred)
if test_pred_tta is None:
test_pred_tta = test_pred
else:
test_pred_tta += test_pred
return test_pred_tta
步骤5:迭代训练和验证模型
model = SVHN_Model1()
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), 0.001)
best_loss = 1000.0
use_cuda = False
if use_cuda:
model = model.cuda()
for epoch in range(2):
train_loss = train(train_loader, model, criterion, optimizer, epoch)
val_loss = validate(val_loader, model, criterion)
val_label = [''.join(map(str, x)) for x in val_loader.dataset.img_label]
val_predict_label = predict(val_loader, model, 1)
val_predict_label = np.vstack([
val_predict_label[:, :11].argmax(1),
val_predict_label[:, 11:22].argmax(1),
val_predict_label[:, 22:33].argmax(1),
val_predict_label[:, 33:44].argmax(1),
val_predict_label[:, 44:55].argmax(1),
]).T
val_label_pred = []
for x in val_predict_label:
val_label_pred.append(''.join(map(str, x[x!=10])))
val_char_acc = np.mean(np.array(val_label_pred) == np.array(val_label))
print('Epoch: {0}, Train loss: {1} \t Val loss: {2}'.format(epoch, train_loss, val_loss))
print(val_char_acc)
# 记录下验证集精度
if val_loss < best_loss:
best_loss = val_loss
torch.save(model.state_dict(), './model.pt')
步骤6:对测试集样本进行预测,生成提交文件
test_path = glob.glob('../input/test_a/*.png')
test_path.sort()
test_label = [[1]] * len(test_path)
print(len(val_path), len(val_label))
test_loader = torch.utils.data.DataLoader(
SVHNDataset(test_path, test_label,
transforms.Compose([
transforms.Resize((64, 128)),
transforms.RandomCrop((60, 120)),
# transforms.ColorJitter(0.3, 0.3, 0.2),
# transforms.RandomRotation(5),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])),
batch_size=40,
shuffle=False,
num_workers=10,
)
test_predict_label = predict(test_loader, model, 1)
test_label = [''.join(map(str, x)) for x in test_loader.dataset.img_label]
test_predict_label = np.vstack([
test_predict_label[:, :11].argmax(1),
test_predict_label[:, 11:22].argmax(1),
test_predict_label[:, 22:33].argmax(1),
test_predict_label[:, 33:44].argmax(1),
test_predict_label[:, 44:55].argmax(1),
]).T
test_label_pred = []
for x in test_predict_label:
test_label_pred.append(''.join(map(str, x[x!=10])))
import pandas as pd
df_submit = pd.read_csv('../input/test_A_sample_submit.csv')
df_submit['file_code'] = test_label_pred
df_submit.to_csv('renset18.csv', index=None)
赛题理解
赛题数据
赛题以街道字符为为赛题数据,数据集报名后可见并可下载,该数据来自收集的SVHN街道字符,并进行了匿名采样处理。
训练集数据包括3W张照片,验证集数据包括1W张照片,每张照片包括颜色图像和对应的编码类别和具体位置;为了保证比赛的公平性,测试集A包括4W张照片,测试集B包括4W张照片。
数据标签
对于训练数据每张图片将给出对于的编码标签,和具体的字符框的位置(训练集、验证集都给出字符位置),可用于模型训练:
Field Description
top 左上角坐标X
height 字符高度
left 左上角坐标Y
width 字符宽度
label 字符编码
字符的坐标具体如下所示:
在比赛数据(训练集和验证集)中,同一张图片中可能包括一个或者多个字符,因此在比赛数据的JSON标注中,会有两个字符的边框信息:
评测指标
选手提交结果与实际图片的编码进行对比,以编码整体识别准确率为评价指标。任何一个字符错误都为错误,最终评测指标结果越大越好,具体计算公式如下:
Score=编码识别正确的数量/测试集图片数量
读取数据
JSON中标签的读取方式:
import json
train_json = json.load(open('../input/train.json'))
# 数据标注处理
def parse_json(d):
arr = np.array([
d['top'], d['height'], d['left'], d['width'], d['label']
])
arr = arr.astype(int)
return arr
img = cv2.imread('../input/train/000000.png')
arr = parse_json(train_json['000000.png'])
plt.figure(figsize=(10, 10))
plt.subplot(1, arr.shape[1]+1, 1)
plt.imshow(img)
plt.xticks([]); plt.yticks([])
for idx in range(arr.shape[1]):
plt.subplot(1, arr.shape[1]+1, idx+2)
plt.imshow(img[arr[0, idx]:arr[0, idx]+arr[1, idx],arr[2, idx]:arr[2, idx]+arr[3, idx]])
plt.title(arr[4, idx])
plt.xticks([]); plt.yticks([])
解题思路
赛题思路分析:赛题本质是分类问题,需要对图片的字符进行识别。但赛题给定的数据图片中不同图片中包含的字符数量不等,如下图所示。有的图片的字符个数为2,有的图片字符个数为3,有的图片字符个数为4。
字符属性 | 图片 |
---|---|
字符:42 字符个数:2 | |
字符:241 字符个数:3 | |
字符:7358 字符个数:4 |
因此本次赛题的难点是需要对不定长的字符进行识别,与传统的图像分类任务有所不同。
- 简单入门思路:定长字符识别
可以将赛题抽象为一个定长字符识别问题,在赛题数据集中大部分图像中字符个数为2-4个,最多的字符 个数为6个。
因此可以对于所有的图像都抽象为6个字符的识别问题,字符23填充为23XXXX,字符231填充为231XXX。
经过填充之后,原始的赛题可以简化了6个字符的分类问题。在每个字符的分类中会进行11个类别的分类,假如分类为填充字符,则表明该字符为空。
- 专业字符识别思路:不定长字符识别
在字符识别研究中,有特定的方法来解决此种不定长的字符识别问题,比较典型的有CRNN字符识别模型。
在本次赛题中给定的图像数据都比较规整,可以视为一个单词或者一个句子。
- 专业分类思路:检测再识别
在赛题数据中已经给出了训练集、验证集中所有图片中字符的位置,因此可以首先将字符的位置进行识别,利用物体检测的思路完成。
此种思路需要参赛选手构建字符检测模型,对测试集中的字符进行识别。选手可以参考物体检测模型SSD或者YOLO来完成。