ocrweb - OCR service online

Build and deploy the service

The core OCR functionaliy a python wrapper around tessearct. The functionality is made available from a fastAPI python web service calling tesseract. Tesseract and fastAPI application are embedded in a docker image that you can build anr run with:

# build the docker image
./scripts/docker.build

# run the docker image
./scripts/docker.run

To declare the OCR service as a regular linux service on a Debian GNU/Linux 11 (bullseye) platform, do

cd ./system
systemctl enable $(pwd)/ocrweb.service
systemctl start ocrweb.service

The home page of the service is then available on http://localhost:8888/

You can then start, stop or look at the service status with

systemctl <action> ocrweb.service
# where <action> is either start, restart, stop or status

To config a caddy service that will make the ocrweb service available to the external world via http and https on default ports, do the following

cp ./system/CaddyFile /etc/caddy
systemctl start caddy
# use command with start, restart, stop or status according to your needs

See also

systemctl enable <servicename.service>
systemctl daemon-reload

Test the service

The service can be tested with tools provided in client directory. client.py calls the OCR service with N threads and computes the average time for the processing of one image. The test images are in ./client/images directory. The output of the OCR processes is stored in the ./client/output directory (not under git).

cd client
python client.py