Sparse Neural Generative Inference Based Pose Estimation

:page_facing_up: report

Advanced Computer Vision - Fall 2020 - Professor David Fouhey

We envisioned a light weight pose estimation model that learns end-to-end to leverage moving particles that sample portions of the image to inform object pose estimation.

The image shows the produced pipeline. Movitated by reducing model and time coplexity, we aimed to utilize particle-filter like sampling to attend to areas where objects might be present in high resolution images.