WPI-Computer Science Department, MS Thesis Presentation Ashkan Ganj , " Robust Depth Estimation via Monocular and Focal Stack Data Fusion for Mobile AR"

 Robust Depth Estimation via Monocular and Focal Stack Data Fusion for Mobile AR

Ashkan Ganj 

PhD 90 student

WPI – Computer Science Department 

Monday, December 9, 2024

Time: 11:00 a.m. – 12:00 p.m.

Location: Fuller Labs 311

 Advisor: Prof. Tian Guo, WPI – Computer Science Department 

Reader: Prof. Jun Dai, WPI – Computer Science Department 

Abstract:

Metric depth estimation is crucial for mobile AR, enabling realistic interactions like object placement and occlusion rendering. While LiDAR offers accurate depth, its limited availability on high-end devices and environmental sensitivity makes it less ideal. Monocular depth estimation, using ubiquitous mobile cameras, is a promising alternative.

We propose HybridDepth a robust monocular depth estimation pipeline that tackles challenges like scale ambiguity, hardware heterogeneity, and generalizability. It utilizes focal stacks, readily available on mobile devices, combined with depth priors from recent single-image depth estimation advancements to produce accurate metric depth maps.