Skip to content

Latest commit

 

History

History
128 lines (81 loc) · 3.01 KB

README.md

File metadata and controls

128 lines (81 loc) · 3.01 KB
Raw

OCR-Captcha-Cracker

A Wrap up tutorial for building OCR module based on CRNN+CTC structure and PyTorch framework

Install

Please clone this repo and install by using following command:

git clone https://github.com/jeff52415/OCR-Captcha-Cracker.git
cd OCR-Captcha-Cracker

## OSX / Linux
pip install -e .["torch"]

## Windows
pip install -e .
pip install torch==1.6.0+cpu torchvision==0.7.0+cpu -f https://download.pytorch.org/whl/torch_stable.html

To install other development dependencies, you need to use this command:

pip install -e .["dev"]

If you are using Windows, please install with following command:

pip install -e .["dev"] -f https://download.pytorch.org/whl/torch_stable.html

Information

Module pretrained
CaptchaCracker Pretrained

General Functions

  • .fit: training function
  • .process : inference function
  • .save : Save current checkpoint to disk

Inference

import string
import random
from captcha.image import ImageCaptcha
from captchacracker.model import CaptchaCracker
model = CaptchaCracker(weight_path='weights/light.pth', backbone='light')


characters = string.digits + string.ascii_uppercase
width, height, n_len, n_class = 128, 64, 4, len(characters)
generator = ImageCaptcha(width=width, height=height)
random_str = ''.join([random.choice(characters) for j in range(n_len)])
img = generator.generate_image(random_str)

output = model.process(img)

Train

from captchacracker.model import CaptchaCracker

model = CaptchaCracker(weight_path='weights/light.pth')

model.fit(batch_size=32, max_iterations=1000000, iteration_per_epoch=2000, save_path='lighter_backbone_pretrained/crnn_ctc_model.pth')

Serving

To serving the model, please use this command:

python captchacracker/serving/serving.py

This command will run a Flask server to serving your model. After the server is ready, you can use curl command to use this model:

curl -X POST 'http://localhost:5000' \
  -F 'image=@tests/assets/test.png'

The response should be:

{
  "result": "3CN"
}

Docker

To deploy this model, we suggest to use Docker to help you simplify the procedure.

You need to make sure you already setup the Docker on your machince. Please check following link to install the Docker:

After the docker is ready, you can use the build script to build the image with your model.

Then, after your image is ready, you can use the run script to run the container and serving your model on any machine.