This repository contains the code for "TriStereoNet: A Trinocular Framework for Multi-baseline Disparity Estimation" [arXiv] [project page][dataset].
TriStereoNet is a multi-baseline solution for depth estimation of both short- and long-range objects. The network is an end-to-end model that fuses the data of narrow and wide stereo pairs. By providing improved flexibility and scalability in depth estimation, this multi-baseline model is promising for long-range visual perception and autonomous navigation, where the image content is unpredictable in terms of the distance to the camera.
| Self-supervised Learning Trinocular EISATS Dataset |
Supervised Learning Trinocular CARLA Dataset |
|---|---|
 |
 |
| Supervised Learning Binocular KITTI 2015 Dataset |
|---|
  |
  |
  |
To be updated ...