Timezone: »

Learning a Multi-View Stereo Machine
Abhishek Kar · Christian Häne · Jitendra Malik

Mon Dec 04 06:30 PM -- 10:30 PM (PST) @ Pacific Ballroom #94 #None

We present a learnt system for multi-view stereopsis. In contrast to recent learning based methods for 3D reconstruction, we leverage the underlying 3D geometry of the problem through feature projection and unprojection along viewing rays. By formulating these operations in a differentiable manner, we are able to learn the system end-to-end for the task of metric 3D reconstruction. End-to-end learning allows us to jointly reason about shape priors while conforming to geometric constraints, enabling reconstruction from much fewer images (even a single image) than required by classical approaches as well as completion of unseen surfaces. We thoroughly evaluate our approach on the ShapeNet dataset and demonstrate the benefits over classical approaches and recent learning based methods.

Author Information

Abhishek Kar (UC Berkeley)

Abhishek Kar is a 5th year graduate student in Jitendra Malik’s lab at UC Berkeley. He received his B.Tech in Computer Science from IIT Kanpur in 2012. Abhishek is the recipient of the CVPR Best Student Paper award in 2015 for his work on category specific shape reconstruction. His research interests lie in 3D computer vision, deep learning and computational photography. He has also spent time at Microsoft Research working on viewing large imagery on mobile devices and at Fyusion capturing "3D photos" with mobile devices and developing deep learning models for them. Some features he has shipped/worked on at Fyusion include 3D visual search, creation of user generated AR/VR content and real-time style transfer on mobile devices.

Christian Häne (UC Berkeley)
Jitendra Malik (University of California at Berkley)

More from the Same Authors