This week we have begun our live testing process. This process collects a new dataset to be used in optimizing our models. While collecting new data, our latest models are run on each collected frame to immediately gauge our performance. In four days, we have collected new data and conducted preliminary testing on eight plugs. The process of this live testing and data collection is significantly faster than our previous data collection for training since we do not have to collect as many images.

Our current performance is within our expectations. Performance between plugs and pins vary sometimes significantly. The next steps in improving our models will be in thorough investigation of our deep learning models. This consists of analyzing outputs of specific frames to try to gain an understanding of how our black-box model works.