README.md

---
license: bsd-3-clause
task_categories:
- depth-estimation
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
dataset_info:
  features:
  - name: image.png
    dtype: image
  - name: depth_1.png
    dtype: image
  - name: depth_2.png
    dtype: image
  - name: depth_3.png
    dtype: image
  - name: depth_4.png
    dtype: image
  - name: depth_5.png
    dtype: image
  - name: depth_6.png
    dtype: image
  - name: depth_7.png
    dtype: image
  - name: depth_8.png
    dtype: image
  - name: __key__
    dtype: string
  splits:
  - name: train
    num_bytes: 27507736943.0
    num_examples: 14800
  - name: validation
    num_bytes: 914329597.0
    num_examples: 500
  download_size: 24747583982
  dataset_size: 28422066540.0
---

LayeredDepth-Syn is a synthetic dataset for [Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation](https://arxiv.org/abs/2503.11633) . If this dataset was helpful for your research, please cite

```
@misc{wen2025seeingseeingglassreal,
      title={Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation}, 
      author={Hongyu Wen and Yiming Zuo and Venkat Subramanian and Patrick Chen and Jia Deng},
      year={2025},
      eprint={2503.11633},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2503.11633}, 
}
```