Low-Level Vision

Project Overview

We have worked on basic algorithmic building blocks that support visual data processing. Our algorithms have advanced the state of the art in optical flow estimation, visual odometry, bottom-up segmentation, intrinsic image decomposition, and other low-level vision problems.


Monocular Visual-Inertial Depth Estimation

Shape from Polarization for Complex Scenes in the Wild

High-dimensional Convolutional Networks for Geometric Pattern Recognition

Does Computer Vision Matter for Action?

Connecting the Dots: Learning Representations for Active Monocular Depth Estimation

Motion Perception in Reinforcement Learning with Dynamic Objects

Deep Fundamental Matrix Estimation

Interactive Image Segmentation with Latent Diversity

Playing for Benchmarks

Accurate Optical Flow via Direct Cost Volume Processing

Direct Sparse Odometry

Full Flow: Optical Flow Estimation By Global Optimization over Regular Grids

Learning to Propose Objects

Geodesic Object Proposals

Fast MRF Optimization with Application to Depth Reconstruction

A Simple Model for Intrinsic Image Decomposition with Depth Cues

Efficient Nonlocal Regularization for Optical Flow