Title: Multi-view NRSfM: Affordable setup for high-fidelity 3D reconstruction

Abstract: Triangulating a point in 3D space should only require two

corresponding camera projections. However in practice, expensive

multi-view setups — involving tens sometimes hundreds of cameras —

are required to obtain the high fidelity 3D reconstructions

necessary for many modern applications. In this talk, we argue that

similar fidelity can be obtained using as little as two cameras by

breaking the tenet of rigidity which is central to much of modern

multi-view geometry. Our approach instead leverages recent advances in

Non-Rigid Structure from Motion (NRSfM) using neural shape priors

while also enforcing multi-view equivariance. We

show how our method can achieve comparable fidelity to expensive

multi-view rigs using only two physical camera views.

Committee:

Prof. Simon Lucey (advisor)

Prof. Laszlo Jeni (co-advisor)

Prof. Katerina Fragkiadaki

Nathaniel Chodosh

Location (Zoom): https://cmu.zoom.us/j/6542017263?pwd=eFNPdVlUVk9PbFUxTEo1QmlWY0Rqdz09

Meeting ID: 654 201 7263

Passcode: 937231