This was a proof of concept developed for a Japanese multinational automobile manufacturer. This project aimed to extract relevant data from a diverse set of engineering drawings using techniques like template matching and regular expressions. Our biggest challenge was that we couldn’t use the coordinates to extract the fields because the drawings were not in the same format.
My Role and Responsibilities
- I thoroughly analyzed the drawings with my teammate to understand the commonalities between the drawings and the rules we could use.
- Used template matching with OCR to extract the fields we needed.
- OCR was not always reliable so we used regular expressions to check if the extracted field was in the format that we wanted.
Libraries
- OpenCV
- Pytesseract
- regex
Conclusion
From the given drawings we were able to extract all the required fields.