The team has finished up data collection, completing the second set of data to be used in training and testing. With that being completed, some preliminary testing has been done with the feature extraction approach for the pin classification. Before putting together both the feature extraction and pin maps that were created, an additional step was done as well which is probe tracking to allow the system to identify changes in position of the probe overtime. For the images above, the left image was able to correctly identify the position of the probe and relate that to the pin map for a correct prediction. The right had some slight confusion with the position of the probe being close to two different pins which is a problem the team is now currently working on for feature extraction.

With the completion of the dataset, the deep learning approaches for both the pin and plug classification has moved into the stage of fine tuning the models for classification. The team plans to run live test and continually improve different parameters and other optimizations that can be done during the next few weeks.