Projects
Neighborhood Matters: Clustering-Driven Regression for House Price Prediction
This project made use of clustering techniques to analyze neighborhood data and predict house prices. This was probably my favorite project of all because it was a novel approach that worked well across a wide range of models, and has applications outside of predicting house prices.
No Frame Left Behind: Repetitive Action Counting with State Space Transformers
This project combined a CNN for feature extraction, a temporal self-similarity matrix for temporal analysis, and a state space transformer for action counting. It was a lot of fun to work on and I learned a lot about how to combine different model architectures.
Beyond the Horizon: Camera Pose and 3D Scene Reconstruction from Single-View Geometry
This project was a lot of fun because it involved no deep learning, just some good old fashioned geometry and linear algebra. The goal was to determine the camera position and reconstruct a 3D scene from a single image using a set of known reference measurements
Reading the Market: Sentiment Analysis of Analyst Ratings for Stock Price Prediction
This project made use of the BERT tranformer architecture and a dataset that was gathered by yours truly from the Yahoo Finance API.
Reviving Dead Pixels: CNN-Based Image Reconstruction
This was a simple project that used a custom convolutional neural network made in PyTorch to fill in missing pixels in images.