This week, we have been making progress in each area of the project and beginning to merge some areas.
Mae worked on determining the accuracy of the yolov5 model, as well as the possibility of introducing inference learning if real-time inference was inaccurate. Mae and Serena worked on real-time inference, including integrating the trained yolov5 model with our hardware by importing the trained model to the Jetson and creating a simple script to test out the inference model on actual plastic bottles. Serena then modified the model script as well as the openCV tracking script to include LiDAR depth data on the bounding boxes that were drawn and labeled by the yolov5 model.
Meghana worked on finalizing the parts list and Serena placed all the orders for our current parts list. Finally, Meghana also wrote initial code for controlling the iRobot through UART by using a Python wrapper library, grating us full mobility of the robot. She also modified our robot intake design by adding wheels to the front of the intake to provide more support.
We believe we are reasonably on schedule, with all parts ordered and software components coming together.
0 Comments