Improvements have been made with deep learning model training workflow. Given that we have four separate datasets available to us, we now can declare which datasets to use for each of the train, validation, and testing sets for the model. As well, we can concatenate multiple datasets together. This gives us greater flexibility in how we train our models. This is the first stage of our final tuning of our deep learning models for plug classification. Pin classification using a deep neural network has been archived as we have greater faith in our hand-written feature extraction approach. However, the pin classification models we have been creating will still be tested.

Live testing will start next week for plug and pin classification using both deep learning and feature extraction models. This is impactful for the deep learning models as it gives us true insight to model performance. Using this new live testing dataset, we can perform final tuning. We are also excited for our QRB 2 next Tuesday and what suggestions the review board has for us.