Deep Learning

In this project, we implemented a face detection model using Detectron2 and several different machine learning algorithms for the facial expression recognition. Our models are trained on two separate image based data sets. We were able to achieve a 70% AP50 with our face detection model and our best facial expression recognition model achieved an accuracy of 65%.

In this project, we implemented several classical machine learning algorithms to predict the severity of car accidents in the US. We also used LSTM, a special recurrent neural net, to analyze the accident description and predict the accident severity. We achieved a best accuracy of 92%.