Model Design: Tested R-CNN, SSD, HOG, and YOLO models for airplane detection in high-resolution satellite imagery. Synthesized the models' low-level techniques that are suitable for satellite image domain
•
Optimization: Reduced the proposal regions by 95% while maintaining comparable model performance quality to shorten experimental times by using selective search, edge detection, and image analysis
Built software to automate removing watermarks from 10,500 mammography images.
•
ML Model Design: Redesigned state-of-the-art breast cancer diagnosis Deep Learning architecture, improving the benchmark AUC by 7%. Wrote PyTorch script and trained models on Google Cloud Platform
