Awesome

Detect, Replace, Refine: Deep Structured Prediction For Pixel Wise Labeling

This repository implements the paper "Detect, Replace, Refine: Deep Structured Prediction For Pixel Wise Labeling" that was accepted in CVPR 2017. An extended technical report is uploaded here: https://arxiv.org/abs/1612.04770

Download pre-computed disparity results

We provide pre-computed disparity maps generated by the "Detect + Replace + Refine x2" architecture. The disparity maps are saved as uint16 PNG images following the instructions of KITTI 2015 dev kit.

KITTI 2015 disparity results:

• Train (200 images)(80.9 MB)
• Test (200 images)(80.7 MB)

The code will soon be available.