Akash Bapat


At Uxmal pyramid, Mayan architecture.

View My GitHub Profile

Akash Bapat | Computer Vision

Hi, my name is Akash Bapat. I am a Ph.D student in the computer science department of UNC Chapel Hill advised by Jan-Michael Frahm. My interests are computer vision and augmented and virtual reality. My research interests can be traced to my IITGN days where I worked with Prof. Raman.

Recent News


Boundary-aware 3D Building Reconstruction from a Single Overhead Image
Jisan Mahmud, True Price, Akash Bapat, and Jan-Michael Frahm
To appear in CVPR, 2020

Predicts building boundaries, semantics, signed distance function (BPSH) and normalized DSMs using a single overhead or satellite image using a multi-task deep learning framework.

Mapped Convolutions
Marc Eder, True Price, Thanh Vu, Akash Bapat, and Jan-Michael Frahm

Decouples weighted sum and sampling in a convolution operation to enable processing 360 imagery which show high levels of distortion.

The Domain Transform Solver
Akash Bapat and Jan-Michael Frahm
CVPR, 2019
arxiv / code / bibtex

Fast edge-aware optimization can be done by using approximate 1-D filtering techniques.

Rolling Shutter and Radial Distortion are Features for High Frame Rate Multi-camera Tracking
Akash Bapat, True Price and Jan-Michael Frahm
CVPR, 2018
poster / supp / bibtex

Radial distortion induces multiple virtual rolling shutter cameras. Using these virtual cameras, we can better constrain the head-pose motion and still track at a high-frequency.

Towards Kilo-Hertz 6-DoF Visual Tracking Using an Egocentric Cluster of Rolling Shutter Cameras
Akash Bapat, Enrique Dunn and Jan-Michael Frahm
ISMAR/TVCG, 2016, Best Paper Award
talk / bibtex

Rolling shutter exposure provides us with a high fequency of row-image samples. If we can estimate a pose per-row, then we have a high-frequency tracker.