This project aims to implement a computer vision system for evaluating thin films produced by a solution-based process. All processes are conducted in an autonomous lab, with experimental procedures carried out by robotic arms or automated equipment. All image sets of the films are captured by an RGBD camera mounted above the robot arm. Using convolutional neural networks, we develop deep learning models for binary classification to differentiate between fabrication success and failure. I analyze the correlation between process variables and fabrication success and identify the critical variables in the fabrication process.
[1] https://github.com/killnice/yolov5-D435i
[2] https://github.com/ultralytics/yolov5
[3] https://github.com/ac-rad/MVTrans?tab=readme-ov-file