Underwater object detection for plastics and fishes

1. Client's requirement: He was working on an Underwater ROV that would detect and collect underwater wastes. To do this, They needed A CNN model which could be deployed in real-time ( on a raspberry pi).

2. Contribution: I collected all the images ( around 5k ), and labeled and annotated them. I tested the models they provided and gave them a report of each model. I also suggested a few configurations and models that would work well with their project.

3. Results: The software part was a big success and was working in real-time with around 13 fps on raspberry pi.