Analysis and Comparison of generative models for Optical Flow Estimation

Advisor: Prof. Amir Barati Farimani, Mechanical Engineering, Carnegie Mellon University

Network Architecture for GAN based Optical Flow Estimator

• Designed architectures for GAN and VAE based optical flow estimators with an image pair conditioned generator.

• Compiled a comparative study of the above methods with the commonly used auto-encoder based optical flow estimators

• The project report can be downloaded from here.