Optical-Flow-Estimation
Tayyabah-Rehman/Optical-Flow-Estimation
Jupyter Notebook
A comparative study of optical flow estimation methods: Lucas-Kanade (sparse) vs RAFT (dense deep learning). Lucas-Kanade tracks 88 feature points with ~1.52px flow. RAFT computes 230K per-pixel vectors capturing complex motions, occlusions, and subtle movements.