Deployed application: imagetotext.app/
Text extraction is based on hint: How to extract text from image in Python If you have any questions, please ask in comments under hint.
How to self-host it
This repo holds a Dockerfile with all needed (Correct version of Python and all dependencies which are managed via Pipenv).
Example of instantiation, e.g. Docker Compose:
services:
imagetotext:
network_mode: host
build: <path_to_this_repo>
command: /bin/bash -c "cd /code/ && pipenv run uvicorn main:app --reload --host 0.0.0.0 --port 8314 --workers 6"
restart: always
Then you can use http://<ip_of_server>:8314 or use any proxy server to route requests to port 8314, e.g. Nginx:
# imagetotext.app
server {
server_name imagetotext.app;
listen 80;
listen 443 ssl;
ssl_certificate /etc/ssl/cert/$CERT_NAME.crt;
ssl_certificate_key /etc/ssl/cert/$CERT_NAME.key;
charset utf-8;
client_max_body_size 75M;
location / {
proxy_set_header Authorization "";
proxy_connect_timeout 400s;
proxy_read_timeout 400s;
proxy_set_header X-Forwarded-For $$proxy_add_x_forwarded_for;
proxy_set_header Host $$http_host;
proxy_redirect off;
proxy_buffers 128 8k;
proxy_buffer_size 16k;
proxy_pass http://127.0.0.1:8314;
add_header Last-Modified $$date_gmt;
etag off;
}
}
Run dev autorelaod server locally
Tested on Ubuntu 20.04 (Native and WSL 2):
1. Install deps
apt install tesseract-ocr libtesseract-dev libleptonica-dev pkg-config
⚠ Do this beforepipenv sync
Tesseract should have version 4.1.1+.
2. Then enter a repo dir and do:
pipenv sync
3. Run app:
pipenv run uvicorn main:app --reload
Testing and troubleshooting
curl -d '{"base64":"baeldung"}' -H 'Content-Type: application/json' http://127.0.0.1:8000/upload/
Or paste same JSON in testreq.json file and run:
curl -d @testreq.json -H 'Content-Type: application/json' http://127.0.0.1:8000/upload/
To test only tesseract on your computer:
tesseract `absolute path to any image with text (/home/ykorolikhin/Pictures/test_text.png)` stdout