---
tags:
- depth
- intrinsics
- image
- geometry
size_categories:
- 1K<n<10K
---


Processed versions of some open-source datasets for evaluation of monocular geometry estimation.

| Dataset | Source | Publication | Num images | Storage Size | Note
| --- | --- | --- | --- | --- | --- |
| NYUv2 | [NYU Depth Dataset V2](https://cs.nyu.edu/~fergus/datasets/nyu_depth_v2.html) | [1] | 654 | 243 MB | Offical test split. Mirror, glass and window manually removed. Depth beyound 5 m truncated. |
| KITTI | [KITTI Vision Benchmark Suite](https://www.cvlibs.net/datasets/kitti/) | [2, 3] | 652 | 246 MB | Eigen's test split. |
| ETH3D | [ETH3D SLAM & Stereo Benchmarks](https://www.eth3d.net/) | [4] | 454 | 1.3 GB | Downsized from 6202×4135 to 2048×1365 |
| iBims-1 | [iBims-1 (independent Benchmark images and matched scans - version 1)](https://www.asg.ed.tum.de/lmf/ibims1/) | [5, 6] | 100 | 41 MB |
| GSO | [Google Scanned Objects](https://research.google/blog/scanned-objects-by-google-research-a-dataset-of-3d-scanned-common-household-items/) | [7] | 1030 | 94 MB | Rendered images at 512x512 resolution.
| Sintel | [MPI Sintel Dataset](http://sintel.is.tue.mpg.de/) | [8] | 1064 | 530 MB | Sky regions manually removed.
| DDAD | [DDAD - Dense Depth for Autonomous Driving](https://github.com/TRI-ML/DDAD) | [9] | 1000 | 593 MB | Uniformly selected from the original validation set. The vechile itself occupies a small portion of the image, which is cropped out.
| DIODE | [DIODE: A Dense Indoor and Outdoor DEpth Dataset](https://diode-dataset.org/) | [10] | 771 | 558 MB | Validation split, 325 indoor and 458 outdoor images. Boundary artifacts filtered. FOV corrected as addressed in this [issue](https://github.com/diode-dataset/diode-devkit/issues/20) |
| Spring | [Spring dataset and benchmark](https://spring-benchmark.org/) | [11] | 1000 | 1.6 GB | Uniformly selected 1/5 of the frames from the original 5000 frames. |
| HAMMER | [HAMMER - Highly Accurate Multi-Modal Dataset for DEnse 3D Scene Regression](https://www.campar.in.tum.de/public_datasets/2022_arxiv_jung/) | [12] | 775 | 397 MB | 

## Download

Make sure `huggingface_hub` is installed. If not installed:

```bash
pip install huggingface_hub
```

Download:

```bash
huggingface-cli download Ruicheng/monocular-geometry-evaluation --repo-type dataset --local-dir /PATH/TO/DOWNLOAD/DIRECTORY --local-dir-use-symlinks False
```

Extract the zip files: (each dataset has a separate zip file)

```bash
cd /PATH/TO/DOWNLOAD/DIRECTORY
unzip '*.zip'
```

Please refer to [https://github.com/microsoft/MoGe/blob/main/moge/utils/io.py](https://github.com/microsoft/MoGe/blob/42cb86af5e0f8873ff966f758033fac30a86fa49/moge/utils/io.py#L86) for reading the depth maps.


## Reference

[1] Silberman, N., Hoiem, D., Kohli, P. and Fergus, R., 2012. Indoor segmentation and support inference from rgbd images. In Computer Vision–ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part V 12 (pp. 746-760). Springer Berlin Heidelberg. 

[2] Geiger, A., Lenz, P. and Urtasun, R., 2012, June. Are we ready for autonomous driving? the kitti vision benchmark suite. In 2012 IEEE conference on computer vision and pattern recognition (pp. 3354-3361). IEEE. 

[3] Geiger, A., Lenz, P., Stiller, C. and Urtasun, R., 2013. Vision meets robotics: The kitti dataset. The international journal of robotics research, 32(11), pp.1231-1237.

[4] T. Schöps, T. Sattler, M. Pollefeys, "BAD SLAM: Bundle Adjusted Direct RGB-D SLAM", Conference on Computer Vision and Pattern Recognition (CVPR), 2019.

[5] Koch, Tobias; Liebel, Lukas; Körner, Marco; Fraundorfer, Friedrich: Comparison of monocular depth estimation methods using geometrically relevant metrics on the IBims-1 dataset. Computer Vision and Image Understanding (CVIU) 191, 2020, 102877

[6] Koch, Tobias; Liebel, Lukas; Fraundorfer, Friedrich; Körner, Marco: Evaluation of CNN-Based Single-Image Depth Estimation Methods. Proceedings of the European Conference on Computer Vision Workshops (ECCV-WS), Springer International Publishing, 2019, 331-348

[7] Downs, L., Francis, A., Koenig, N., Kinman, B., Hickman, R., Reymann, K., McHugh, T.B. and Vanhoucke, V., 2022, May. Google scanned objects: A high-quality dataset of 3d scanned household items. In 2022 International Conference on Robotics and Automation (ICRA) (pp. 2553-2560). IEEE.

[8] Butler, D.J., Wulff, J., Stanley, G.B. and Black, M.J., 2012. A naturalistic open source movie for optical flow evaluation. In Computer Vision–ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part VI 12 (pp. 611-625). Springer Berlin Heidelberg.

[9] Guizilini, V., Ambrus, R., Pillai, S., Raventos, A. and Gaidon, A., 2020. 3d packing for self-supervised monocular depth estimation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 2485-2494).

[10] Vasiljevic, I., Kolkin, N., Zhang, S., Luo, R., Wang, H., Dai, F.Z., Daniele, A.F., Mostajabi, M., Basart, S., Walter, M.R. and Shakhnarovich, G., 2019. Diode: A dense indoor and outdoor depth dataset. arXiv preprint arXiv:1908.00463.

[11] Mehl, L., Schmalfuss, J., Jahedi, A., Nalivayko, Y. and Bruhn, A., 2023. Spring: A high-resolution high-detail dataset and benchmark for scene flow, optical flow and stereo. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 4981-4991).

[12] Jung, H., Ruhkamp, P., Zhai, G., Brasch, N., Li, Y., Verdie, Y., Song, J., Zhou, Y., Armagan, A., Ilic, S. and Leonardis, A., 2023. On the importance of accurate geometry data for dense 3D vision tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 780-791).