GUI takes webcam input and displays it on viewing screen and iterates through steps. A functional override button is included.

Intersection of bounding boxes appears on YOLO detection feed for one object pair

Model trained on small dataset allen key images and hand dataset can identify allen key with 97% surety and hands with over 45% surety. 

The Raspberry Pi camera and Econ E-cam feed now show up together, although frame rate is slower. YOLOv7 is running on both streams. Jetson Orin Nano's power is strained.

The Econ E-Cam now works without fish eye distortion. YOLOv7 can successfully take the live camera stream as input. However, the model trained on the data we collected classifies nearly everything as a monkey wrench!

Data collection rig uses a rotating platform to take 3K images of rotating tools