CICS Team Wins Honorable Mention at CVPR 2018

PhD student Hang Su, along with UMass Amherst College of Information and Computer Sciences Assistant Professors Subhransu Maji and Evangelos Kalogerakis, have been recognized with an Honorable Mention Best Paper award for their paper “SPLATNet: Sparse Lattice Networks for Point Cloud Processing” at the Computer Vision and Pattern Recognition (CVPR 2018) conference in Salt Lake City, June 19-21.

Co-authored with collaborators from NVIDIA and the University of California, Merced, the paper describes an architecture that directly operates on point cloud data acquired from depth sensors such as LiDAR to produce semantic labels. Such techniques are becoming increasingly important for many applications, notably autonomous cars.

Since point clouds consists of unordered set of points, developers must preprocess them into grid-structured data such as voxels in order to apply conventional deep architectures. However, voxel representations creates data with more artifacts and fewer details, an issue researchers have struggled to resolve for years.

“SPLATNet directly takes point clouds as input and computes hierarchical and spatially aware features with sparse and efficient lattice filters,” the researchers wrote in their paper. “SPLATNet allows easy mapping of 2D information into 3D and vice-versa, resulting in a novel network architecture for joint processing of point clouds and multi-view images.”

Su, Maji, and Kalogerakis are members of the CICS Computer Vision Laboratory and are affiliated with the college’s Center for Data Science. Together, they pioneer development of vision systems that can function in difficult environments, applying their research of underlying scientific principles to create real-world solutions.

The code for SPLATNet is available on GitHub.

Congratulations Hang, Subhransu, and Vangelis!