•Contribute towards solving machine perception research problems with deep learning.
•Implemented a new metric for Pedestrian Trajectory Projection that provides Euclidean distance, Bivariate Mahalanobis Distance, and multivariate Mahalanobis distance. The multivariate mahalanobis distance takes in data, fits a Gaussian Mixture Model with the best BIC to find the clusters, and then uses clusters to calculate the distance.
•Implemented custom Pytorch Dataset for time series. Designed a CNN model and implemented an LSTM model for vehicle following. Implemented bivariate negative log loss, average displacement error, and final displacement error metrics.
