My project was focused on image segmentation for use in training semantic segmentation neural networks, ultimately for use in self-driving cars. I took two approaches. First, a purely heuristic approach (written in Matlab), which leveraged LiDAR data to form hierarchical depth clusters, from which objects of interest could be segmented. This approach was extended to incorporate a GUI. Second, a neural network approach (using Tensorflow) which first identified coarse bounding boxes around objects of interest and then refined the bounding boxes by hierarchical clustering of LiDAR data. Click here for a copy of my project.