Bellingham, Washington, United States
• Worked with professor Scott Wehrwein and other students to use semantic features from deep learning semantic segmentation architectures and a small neural network to create a richer notion of pixel similarity than traditional RGB euclidean distance.
• Interfaced with an implementation of the Panoptic-Deeplab panoptic segmentation model from Facebook Research.
• Wrote code in Python using Numpy, OpenCV, and PyTorch.