https://github.com/chapzq77/chinese-ocr

[python3.6] 运用tf实现自然场景文字检测,keras/pytorch实现ctpn+crnn+ctc实现不定长场景文字OCR识别

https://github.com/chapzq77/chinese-ocr

Repository

[python3.6] 运用tf实现自然场景文字检测,keras/pytorch实现ctpn+crnn+ctc实现不定长场景文字OCR识别

https://github.com/chapzq77/CHINESE-OCR/blob/master/

# tensorflowkeras/pytorchOCR

# 

-  090180270 
-  keras keras
- OCR

## 

``` 
Bash
##GPU
sh setup.sh
## CPU
sh setup-cpu.sh
##CPU python3
sh setup-python3.sh

python3.6+tensorflow1.7+cpu/gpu
```

# 
* 3
	 	**1. -Classify(vgg16)**
*  **2. -CTPN(CNN+RNN)**
*  **3. EndToEnd-CRNN(CNN+GRU/LSTM+CTC)**

# -vgg
```bash
VGG16090180270.
angle/predict.py800088.23%
```
[BaiduCloud](https://pan.baidu.com/s/1zquQNdO0MUsLMsuwxbgPYg)

# CTPN

ctpn2016[-TextDetection](http://xiaofengshi.com/2019/01/23/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0-TextDetection/)repoctpnCTPN

![ctpn_model](./asset/ctpn_model.png)

ctpnrepoctpnanchor

```python
def generate_anchors(base_size=16, ratios=[0.5, 1, 2],
                     scales=2 ** np.arange(3, 6)):
    heights = [11, 16, 23, 33, 48, 68, 97, 139, 198, 283]
    widths = [16]
    sizes = []
    for h in heights:
        for w in widths:
            sizes.append((h, w))
    return generate_basic_anchors(sizes)
```

ctpnctpnanchor

# OCR :CRNN

## ocrGRU+CTC

keras pytorchkeraspytorch
- tensorflow[TF:LSTM-CTC_loss](https://github.com/xiaofengShi/CTC_TF) 

## ctc

ctcctpn

seq2seqRNNattention

ctcctcctclabelocrCTCctc[CTC](http://xiaofengshi.com/2019/02/14/CTC%E7%AE%97%E6%B3%95%E5%8E%9F%E7%90%86/)

repoCNN+RNN+CTCCNN+CTCCNNdensenetresnet

# 

## 

```
demo.py  ctpn./ctpn/ctpn/other.py draw_boxescv2.inwrite('dest_path',img)ctpnocr
```
## 

### 1 ctpn
* --./ctpn/ctpn/train_net.py
* vgg[VGG_imagenet.npy](https://pan.baidu.com/s/1JO_ZojA5bkmJZsnxsShgkg)
pretrained_model,
[checkpoint](https://pan.baidu.com/s/1aT-vHgq7nvLy4M_T6SwR1Q)
* ctpn[](https://pan.baidu.com/s/1NXFmdP_OgRF42xfHXUhBHQ)
.ctpn/lib/datasets/pascal_voc.py pascal_voc self.devkit_path

### 2 crnn

* keras ./train/keras_train/train_batch.py  model_path-- 
MODEL_PATH---
[keras](https://pan.baidu.com/s/1vTG6-i_bFMWxQ_7xF06usg)
* pythorch./train/pytorch-train/crnn_main.py
```
parser.add_argument(
    '--crnn',
    help="path to crnn (to continue training)",
    default=)
parser.add_argument(
    '--experiment',
    help='Where to store samples and models',
    default=,)
```
[pytorch](https://pan.baidu.com/s/1LEDNHEr3luloB7eZK6GOeA)


# 
## OCR

![ctpn1](./test/ttttt.png)
`===========================================================`
![ctpn1](./test/test1.png)
`===========================================================`
![ctpn+crnn1](./test/ttttt_result.png)


### 
![ctpn2](./test/test.png)
`===========================================================`
![ctpn2](./test/test_pre.png)
`===========================================================`
![ctpn+crnn2](./test/test_result.png)

# 

### tensorflowcrnntrick(dropuout,bn,learning_decay)
```
crnncrnncnn
rnn+attentionrestnet
rnn+attention+ctc~~~~
```

## 

 fine-tune

### **Chinese Text in the Wild(CTW)**

322851018402(), 2048*204831GB(8:1:1)(25887812872)(3269103519)(3129103519)

```
https://arxiv.org/pdf/1803.00085.pdf 
https://ctwdataset.github.io/
```

### **Reading Chinese Text in the Wild(RCTW-17)**

122638034422911.4GB

```
http://mclab.eic.hust.edu.cn/icdar2017chinese/dataset.html
http://arxiv.org/pdf/1708.09585v2
```

### ICPR MWI 2018 

2000050%50%2GB

```
https://tianchi.aliyun.com/competition/information.htm?raceId=231651&_is_login_redirect=true&accounttraceid=595a06c3-7530-4b8a-ad3d-40165e22dbfe
```

### **Total-Text**

155511459441MB1255 300

```
http://www.cs-chan.com/source/ICDAR2017/totaltext.zip
http:// arxiv.org/pdf/1710.10400v
```

### Google FSNS()

600*15010448681615020404

```
http://rrc.cvc.uab.es/?ch=6&com=downloads
http:// arxiv.org/pdf/1702.03970v1
```

### **COCO-TEXT**

6368617358912.58GB436861000010000

```
http://arxiv.org/pdf/1601.07140v2
https://vision.cornell.edu/se3/coco-text-2/
```

### **Synthetic Data for Text Localisation**

85875072668662897148741GBlabel

```
http://www.robots.ox.ac.uk/~vgg/data/scenetext/
http://www.robots.ox.ac.uk/~ankush/textloc.pdf
Code: https://github.com/ankush-me/SynthText ()
Code https://github.com/wang-tf/Chinese_OCR_synthetic_data()
```

### **Synthetic Word Dataset**

9910GB

```
http://www.robots.ox.ac.uk/~vgg/data/text/
```

### **Caffe-ocr**

360280x3259908.6GB

```
https://pan.baidu.com/s/1dFda6R3
```

## 

- [-OCR_Overview](http://xiaofengshi.com/2019/01/05/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0-OCR_Overview/)
- [pytorch crnn](https://github.com/meijieru/crnn.pytorch.git)    
- [keras-crnn ](https://www.zhihu.com/question/59645822)  
- [tensorflow-crnn](https://github.com/ilovin/lstm_ctc_ocr)
- [tensorflow-ctpn](https://github.com/eragonruan/text-detection-ctpn )
- [CAFFE-CTPN](https://github.com/tianzhi0549/CTPN)

Owner

GitHub Events