Convolutional Connectionist-temporal-classification with Recurrent network(CCR)

• This a project used to do Optical Character Recognition(ocr).

Train & Test

• You can found more information of how to do in Readme.md of "CCR/CCR"

Deploy

Celery

• Celery is an asynchronous task manager.
• Celery support following message proxy(or called "Broker""): RabbitMQ, Redis, MongoDB, Beanstalk, SQLAlchemy, Zookeeper
• RabbitMQ is the most suggested, and second is Redis. Due to RabbitMQ is more difficult to build, so we use Redis as Broker

# Install celery with dependency package
$ pip install "celery[librabbitmq,redis,msgpack]"

• You can start your task manager by the following command:

$ celery worker -A celery_worker.celery -l INFO

Redis

• Install Redis for Celery

$ wget http://download.redis.io/redis-stable.tar.gz
$ tar xvzf redis-stable.tar.gz
$ cd redis-stable
$ make
$ sudo apt-get install tcl8.5-dev
$ make test
$ sudo apt install redis-server

Gunicorn

• Gunicorn is a web server.

$ sudo apt install gunicorn
$ pip install gunicorn
$ pip install gevent

After installed, start your server by running:

$ gunicorn manage:app -c gc_config.py

• File gc_config is used to set parameters for Gunicorn. Format of gc_config is listed as following:

bind = ":5001"
workers = 4
worker_class = "gevent"
errorlog = '-'
loglevel = "info"

Send a request to test server

• Not that This project only doing ocr without any character detect, so you can only send an image focus on target word.
• Data Format

{
	"b64_images": ["YOUR/BASE64/ENCODED/IMAGE/STRING"]
}

Using Postman

• Header: {"content-type": "application/json"}
• Method: POST
• URL: http://localhost:5001/api/v0.0.0/plate-recognition/CCR

Using Command Line

curl -X POST --data YOUR_DATA http://localhost:5001/api/v0.0.0/plate-recognition/CCR

Result

• Server will return {"state": "task accepted"}, if request successful, and you will get result of recognition in celery.