The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Error code: FileFormatMismatchBetweenSplitsError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
PICD-Video CLEVRER Route 2 Processed Pseudo-States
This dataset contains processed CLEVRER Route 2 pseudo-state data used for external video-only stress testing in the PICD-Video project.
PICD-Video studies physics-informed dynamic causal discovery from physical video/object-state sequences. This CLEVRER Route 2 dataset was created by extracting annotation-free object-like pseudo-states from CLEVRER videos and evaluating graph transfer and rollout behavior.
Contents
data/processed/clevrer_route2/- Processed pseudo-states and manifests.
data/processed/clevrer_route2_split/- Train/validation/test split files for fine-tuning experiments.
configs/- Transfer and fine-tuning configs.
results/- Zero-shot and fine-tuning evaluation outputs.
paper_assets/- Paper-ready table files when available.
Preprocessing
The Route 2 preprocessing pipeline performs annotation-free visual pseudo-state extraction from CLEVRER videos. The extracted state representation follows the PICD format:
[x, y, vx, vy, mass, radius]
Pseudo-contact/event labels are derived from extracted object centers and approximate radii.
Important note
This repository contains processed pseudo-state data and evaluation summaries. It does not include raw CLEVRER videos. Users should refer to the original CLEVRER project for the source benchmark.
Code
https://github.com/karthik789338/picd-video
Intended use
- Reproducing PICD-Video CLEVRER Route 2 stress-test results.
- Testing graph recovery from annotation-free video pseudo-states.
- Studying limitations of zero-shot physical-video transfer.
Limitations
CLEVRER Route 2 is treated as an external stress test. In PICD-Video experiments, dynamic graph recovery improves, but trajectory rollout dominance is not claimed due to pseudo-state extraction noise and smooth self-motion bias.
- Downloads last month
- 15