# Characters Recognition
A Chinese characters recognition repository based on convolutional recurrent networks.
<p align='center'>
<img src='images/wechatgroup.png' title='example' style='max-width:600px'></img>
</p>
## Performance
#### Recognize characters in pictures
<p align='center'>
<img src='images/demo.png' title='example' style='max-width:600px'></img>
</p>
<p align='center'>
<img src='images/demo_2.jpg' title='example2' style='max-width:600px'></img>
</p>
## Dev Environments
1. WIN 10 or Ubuntu 16.04
2. **PyTorch 1.2.0 (may fix ctc loss)** with cuda 10.0 ð¥
3. yaml
4. easydict
5. tensorboardX
### Data
#### Synthetic Chinese String Dataset
1. Download the [dataset](https://pan.baidu.com/s/1ufYbnZAZ1q0AlK7yZ08cvQ)
2. Edit **lib/config/360CC_config.yaml** DATA:ROOT to you image path
```angular2html
DATASET:
ROOT: 'to/your/images/path'
```
3. Download the [labels](https://pan.baidu.com/s/1oOKFDt7t0Wg6ew2uZUN9xg) (password: eaqb)
4. Put *char_std_5990.txt* in **lib/dataset/txt/**
5. And put *train.txt* and *test.txt* in **lib/dataset/txt/**
eg. test.txt
```
20456343_4045240981.jpg 89 201 241 178 19 94 19 22 26 656
20457281_3395886438.jpg 120 1061 2 376 78 249 272 272 120 1061
...
```
#### Or your own data
1. Edit **lib/config/OWN_config.yaml** DATA:ROOT to you image path
```angular2html
DATASET:
ROOT: 'to/your/images/path'
```
2. And put your *train_own.txt* and *test_own.txt* in **lib/dataset/txt/**
eg. test_own.txt
```
20456343_4045240981.jpg ä½ å¥½åï¼ç¥å½ï¼
20457281_3395886438.jpg æå®åï¼ä¸çï¼
...
```
**note**: fixed-length training is supported. yet you can modify dataloader to support random length training.
## Train
```angular2html
[run] python train.py --cfg lib/config/360CC_config.yaml
or [run] python train.py --cfg lib/config/OWN_config.yaml
```
```
#### loss curve
```angular2html
[run] cd output/360CC/crnn/xxxx-xx-xx-xx-xx/
[run] tensorboard --logdir log
```
#### loss overview(first epoch)
<center/>
<img src='images/train_loss.png' title='loss1' style='max-width:800px'></img>
</center>
<p>
<img src='images/tb_loss.png' title='loss1' style='max-width:600px'></img>
</p>
## Demo
```angular2html
[run] python demo.py --image_path images/test.png --checkpoint output/checkpoints/mixed_second_finetune_acc_97P7.pth
```
## References
- https://github.com/meijieru/crnn.pytorch
- https://github.com/HRNet
没有合适的资源?快使用搜索试试~ 我知道了~
CRNN_Chinese_Characters_Rec:(CRNN)汉字识别
共44个文件
py:11个
pyc:8个
png:8个
需积分: 49 18 下载量 169 浏览量
2021-02-05
09:52:17
上传
评论 2
收藏 51.36MB ZIP 举报
温馨提示
字符识别 基于卷积递归网络的汉字识别库。 性能 识别图片中的字符 开发环境 WIN 10或Ubuntu 16.04 使用CUDA 10.0的PyTorch 1.2.0(可能会修复CTC丢失) :fire: yaml Easydict 张量板 数据 合成中文字符串数据集 下载 将lib / config / 360CC_config.yaml DATA:ROOT编辑为您的图像路径 DATASET: ROOT: 'to/your/images/path' 下载(密码:eaqb) 将char_std_5990.txt放入lib / dataset / txt / 并将trai
资源详情
资源评论
资源推荐
收起资源包目录
CRNN_Chinese_Characters_Rec-stable.zip (44个子文件)
CRNN_Chinese_Characters_Rec-stable
images
loss_2.png 110KB
wechatgroup.png 113KB
test.png 12KB
wechatpay.jpg 101KB
train_loss.png 180KB
alipay.jpg 112KB
tb_loss.png 30KB
loss_1.png 78KB
demo_2.jpg 42KB
test_2.png 6KB
demo.png 80KB
output
checkpoints
mixed_second_finetune_acc_97P7.pth 44.91MB
360CC
crnn
2020-06-22-15-53
log
events.out.tfevents.1592812406.DESKTOP-LVNAKHJ 25B
.github
FUNDING.yml 827B
README.md 2KB
.idea
crnn.iml 453B
misc.xml 288B
workspace.xml 24KB
vcs.xml 180B
modules.xml 260B
lib
core
function.py 4KB
__pycache__
function.cpython-35.pyc 4KB
dataset
_360cc.py 2KB
txt
test.txt 20.71MB
train.txt 3B
char_std_5990.txt 17KB
__init__.py 243B
__pycache__
_own.cpython-35.pyc 2KB
__init__.cpython-35.pyc 442B
_360cc.cpython-35.pyc 3KB
_own.py 2KB
models
__pycache__
crnn.cpython-35.pyc 4KB
crnn.py 3KB
utils
preprocessing.py 1KB
imutils.py 5KB
utils.py 6KB
__pycache__
utils.cpython-35.pyc 6KB
imutils.cpython-35.pyc 6KB
config
alphabets.py 20KB
OWN_config.yaml 1KB
__pycache__
alphabets.cpython-35.pyc 20KB
360CC_config.yaml 1KB
demo.py 3KB
train.py 5KB
共 44 条
- 1
普通网友
- 粉丝: 24
- 资源: 4570
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功
评论0