Shirley’s Status Report: Week 12

Progress

1. Completed training Yolov3 model for 4 classes, 200 Epochs in total. Accuracy of the trained model is not yet confirmed, will validate later. 
2. Modified CV server to migrate from OpenCV inference to PyTorch inference. Improved style by wrapping up threads in separate classes.
3. Tested the possibility of panorama, reached the conclusion that warping edges do not suit object detection. Tried newly trained model on panorama as well, turned out not working. Decided to focus on deriving the exact angles from a single image first.
4. Discussed and designed the communication scheme between RPi and object detector server. Upon each calibration, RPi needs to confirm with server that no previous job running before setting out to rotate the base. If RPi decides to rotate, then it will be an atomic process until RPi receives the detection result from object detector server.

Schedule

1. Work on data post processing and try to test with camera mounted on robot arm.

Deliverables Next Week

1. A working post-processing equation which calculates the motor position.
2. A full working communication protocol between RPi and  object detector server.