The authors develop a calibration pipeline that combines learned stereo depth with a joint factor graph optimization, pooling all episodes from the same physical robot to recover shared kinematic parameters and per-scene extrinsics.
Depth supervision boosts RGB and geometric consistency in robot world models
A calibration pipeline turns the DROID dataset into metric 3D data, and a Stable Video Diffusion model jointly predicts multi-view RGB and depth, improving RGB by +1.48 dB PSNR.
Academic
Jai Bardhan · Josef Sivic · Vladimir Petrik
Czech Technical University in Prague
Research Digest··2 min read
The authors introduce a calibration pipeline that recalibrates the DROID dataset into a metric 3D corpus (DROID-3D) and train DepthWorld, a video diffusion world model that jointly predicts multi-view RGB and depth.
Why this paper
From Czech Technical University in Prague
In one line
DepthWorld jointly predicts RGB and depth, improving RGB PSNR by 1.48 dB over an RGB-only baseline while providing metric depth.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (2 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§