moaminsharifi/ocr-api

This API uses a pre-trained model from hugging face to convert captcha images into text. It is based on Python and uses FastApi, Docker and Docker Compose in order to provide a simple, fast and efficient way to convert images into text.

OCR - API documentation

This API uses a pre-trained model from hugging face to convert captcha images into text. It is based on Python and uses FastApi, Docker and Docker Compose in order to provide a simple, fast and efficient way to convert images into text.

• OCR - API documentation
  • Setup Project
    • install git lfs (git large file system)
    • Ubuntu
    • update token in .env file
      • go to api/ directory and clone model
      • Build Docker images
      • Run Docker images with docker compose
      • Connect to docker container
    • manage docker containers with GUI
    • project structure example
    • Example Base64 Image

Setup Project

install git lfs (git large file system)

Ubuntu

curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.deb.sh | sudo bash
sudo apt-get install git-lfs

update token in .env file

go to api/ directory and clone model

git clone https://huggingface.co/microsoft/trocr-small-printed

Build Docker images

docker compose build

Run Docker images with docker compose

docker compose up --force-recreate --build

• use -d flag if want to detached after containers up
• use --force-recreate flag for re create docker containers
• use --build flag for re create docker images

For Example ** Build and deployment **

docker compose up -d --build --force-recreate
docker compose --compatibility up -d --build --force-recreate 

Connect to docker container

docker exec -ti ocr-api_ocr-api bash 

use NAME to specify container name, normally docker container name is:

manage docker containers with GUI

Setup portainer docker container

docker run -d  -p 2400:9000 --name portainer --restart=always -v /var/run/docker.sock:/var/run/docker.sock -v portainer_data:/data portainer/portainer-ce:latest

then go to http://localhost:2400 and setup admin credentials

for always started up use this roles:

docker update --restart unless-stopped portainer

project structure example

api
├── sub
│   ├── router.py
│   ├── schemas.py  # pydantic models
│   ├── models.py  # db models
│   ├── dependencies.py
│   ├── config.py  # local configs
│   ├── tasks.py  # celery tasks
│   ├── constants.py
│   ├── exceptions.py
│   ├── service.py
│   └── utils.py
│   │   
│   ├── config.py  # global configs
│   ├── models.py  # global models
│   ├── schemas.py  # pydantic models
│   ├── exceptions.py  # global exceptions
│   ├── pagination.py  # global module e.g. pagination
│   ├── database.py  # db connection related stuff
│   └── main.py
├── .env
|-- .dockerignore
└── .gitignore

Example Base64 Image

import requests

url = "http://localhost:9090/ocr"

payload = {"image": "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"}
headers = {
    "token": "USE_ENV_TOKEN",
    "Content-Type": "application/json"
}

response = requests.request("POST", url, json=payload, headers=headers)

print(response.text)