---
license: cc-by-4.0
task_categories:
- object-detection
tags:
- RGB-D
- video-object-counting
- crowded-scenes
- occlusion
---

# RGBD-VideoCount

RGBD-VideoCount is an RGB-D video dataset for video object counting in crowded and occluded scenes. It provides synchronized RGB frames and depth maps, together with instance-level annotations for evaluating detection, cross-frame association, and video-level de-duplication.

## Dataset Summary

- 195 RGB-D video clips
- 6 object categories
- 2,032 finely annotated frames
- 77,638 instance bounding boxes
- Multi-category shelf and crowded-object scenes
- RGB frames, aligned depth maps, instance annotations, counting annotations, data splits, and visual exemplars

## Directory Structure

```text
RGBD-VideoCount/
|- images/
|- Depth_Data/
|- object_annotations/
|- count_annotations/
|- dataset_split.json
|- video_class.txt
|- exemplars_train.json
|- exemplars_val.json
`- exemplars_test.json
```

## Data Description

- `images/`: RGB video frames.
- `Depth_Data/`: Depth maps aligned with RGB frames.
- `object_annotations/`: Instance-level bounding-box annotations.
- `count_annotations/`: Video-level counting annotations.
- `dataset_split.json`: Training, validation, and test splits.
- `video_class.txt`: Category metadata.
- `exemplars_*.json`: Visual exemplars for exemplar-guided training and evaluation.

## Citation

If you use this code, please cite our paper:

```bibtex
@inproceedings{xu2026depth,
  title     = {Depth-Guided Video Object Counting in Crowded Scenes},
  author    = {Xu, Yuanjing and Liu, Xinyan and Chen, Weidong and Zou, Zixuan and Zhang, Linhao and Meng, Zhuangzhe and Chan, Antoni B. and Zhang, Weigang},
  booktitle = {Proceedings of the 34th ACM International Conference on Multimedia},
  year      = {2026},
  doi       = {10.1145/3767308.3835482}
}
```

## Limitations

RGBD-VideoCount focuses on crowded object scenes and may not represent all real-world environments. Performance can be affected by depth quality, severe appearance ambiguity, camera motion, and unseen object categories. Users are responsible for evaluating suitability before deployment in real applications.

## License

RGBD-VideoCount is released under the [Creative Commons Attribution 4.0 International License](https://creativecommons.org/licenses/by/4.0/). Users must provide appropriate attribution when using, modifying, or redistributing this dataset.

## Related Resources

- Model weights: https://huggingface.co/aerospace123/DG-Net
- Dataset: https://huggingface.co/datasets/aerospace123/RGBD-VideoCount
- Code: https://github.com/streamer-AP/DG-Net