OCR on Product Photos

[cod3monk]

Goal: make items easier findable, without having to manually describe them in detail.

Idea:

1. Run OCR on all photos and extract text
2. Store extracts with photo
3. Extend item search to include photo OCR-extracts

Things to consider:

• Allow this process to be done image-by-image, so that it can be improved in future
• Primary goal would be to run this process in batch, possibly outside of django, but if the implementation is capable of doing the same live on newly uploaded photos this would be a nice feature
• Test cases to determine quality of extracts would be good to have, e.g. compare automatic extract to manual extractions
• Consider comparing multiple OCR systems
• Also extract and store EANs or other barcodes present in photos

[danieloeh]

So far, i have implemented a basic prototype of this feature which lets you run OCR on all images via python manage.py ocr. It uses pytesseract for the OCR. The result is shown below the description of each image in the "Update Item" view.

Feature branch: https://github.com/danieloeh/inventory_management/tree/feature/ocr