Overall, our project is a success. Despite the setbacks caused by COVID-19, we built a software package that is ready to be implemented. Our system is able to accurately recognize authorized Eaton-Bussmann personnel and verify that the correct PPE is present through QR code decoding. We found that the HOG and CNN facial recognition models were equally accurate; however, the HOG model had a consistently lower run time than the CNN model. The completion of our system marks the end of Eaton-Bussmann’s first phase of tackling problems using computer vision. We hope that this project can be used as a building block as future teams tackle larger computer vision problems.
Deliverables
Our initial plan was to deliver a facial recognition and equipment verification system that was part of a hardware system for Eaton-Bussmann to use at their facility. Due to COVID-19, we were unable to build a realized hardware system that was integrated with our software package. Our deliverables are below:
- Comparison of two facial recognition models
- Website that generates QR code images for printing
- Working facial recognition and equipment verification system
- Installation guide that will help Eaton-Bussmann install our software package
- Written Final Report
- Website