File size: 5,562 Bytes
7382c66 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 | # <img src="assets/badges/icon2.png" alt="lotus" style="height:1.2em; vertical-align:bottom;"/> DA<sup>2</sup>: Depth Anything in Any Direction
[](https://depth-any-in-any-dir.github.io/)
[](http://arxiv.org/abs/2509.26618)
[](https://huggingface.co/spaces/haodongli/DA-2)
[](https://huggingface.co/datasets/haodongli/DA-2)
[](https://docs.google.com/presentation/d/1QUonqLuYGEh0qcqY72pbTXsZimINlyN4rOogy7qX4GY/edit?usp=sharing)
[](https://depth-any-in-any-dir.github.io/bibtex.txt)
[Haodong Li](https://haodong2000.github.io/)<sup>123§</sup>,
[Wangguangdong Zheng](https://wangguandongzheng.github.io/)<sup>1</sup>,
[Jing He](https://jingheya.github.io/)<sup>3</sup>,
[Yuhao Liu](https://yuhaoliu7456.github.io/)<sup>1</sup>,
[Xin Lin](https://linxin0.github.io/)<sup>2</sup>,
[Xin Yang](https://abnervictor.github.io/2023/06/12/Academic-Self-Intro.html)<sup>34</sup>,<br>
[Ying-Cong Chen](https://www.yingcong.me/)<sup>34✉</sup>,
[Chunchao Guo]()<sup>1✉</sup>
<span class="author-block"><sup>1</sup>Tencent Hunyuan</span>
<span class="author-block"><sup>2</sup>UC San Diego</span>
<span class="author-block"><sup>3</sup>HKUST(GZ)</span>
<span class="author-block"><sup>4</sup>HKUST</span><br>
<span class="author-block">
<sup>§</sup>Work primarily done during an internship at Tencent Hunyuan.
<sup>✉</sup>Corresponding author.
</span>
<strong>DA<sup>2</sup> predicts dense, scale-invariant distance from a single 360° panorama in an end-to-end manner, with remarkable geometric fidelity and strong zero-shot generalization.</strong>
## 📢 News
- 2025-10-10 The curated panoramic data is released on [huggingface](https://huggingface.co/datasets/haodongli/DA-2)!
- 2025-10-10 The evaluation code and the [testing data](https://huggingface.co/datasets/haodongli/DA-2-Evaluation) are released!
- 2025-10-04 The 🤗Huggingface Gradio demo ([online](https://huggingface.co/spaces/haodongli/DA-2) and [local](https://github.com/EnVision-Research/DA-2?tab=readme-ov-file#-gradio-demo)) are released!
- 2025-10-04 The inference code and the [model](https://huggingface.co/haodongli/DA-2) are released!
- 2025-10-01 [Paper](https://arxiv.org/abs/2509.26618) released on arXiv!
## 🛠️ Setup
> This installation was tested on: Ubuntu 20.04 LTS, Python 3.12, CUDA 12.2, NVIDIA GeForce RTX 3090.
1. Clone the repository:
```
git clone https://github.com/EnVision-Research/DA-2.git
cd DA-2
```
2. Install dependencies using conda:
```
conda create -n da-2 python=3.12 -y
conda activate da-2
pip install -e src
```
> For macOS users: Please remove `xformers==0.0.28.post2` (line 16) from `src/pyproject.toml` before `pip install -e src`, as [xFormers does not support macOS](https://github.com/facebookresearch/xformers/issues/775#issuecomment-1611284979).
## 🤗 Gradio Demo
1. Online demo: [Hugggingface Space](https://huggingface.co/spaces/haodongli/DA-2)
2. Local demo:
```
python app.py
```
## 🕹️ Inference
> We've pre-uploaded the cases appeared in the [project page](https://depth-any-in-any-dir.github.io/). So you can proceed directly to step 3.
1. Images are placed in a directory, e.g., `assets/demos`.
2. (Optional) Masks (e.g., sky masks for outdoor images) in another directory, e.g., `assets/masks`. The filenames under both directories should be consistent.
3. Run the inference command:
```
sh infer.sh
```
4. The visualized distance and normal maps will be saved at `output/infer/vis_all.png`. The projected 3D point clouds will be saved at `output/infer/3dpc`.
## 🚗 Evaluation
1. Download the evaluation datasets from [huggingface](https://huggingface.co/datasets/haodongli/DA-2-Evaluation):
```
cd [YOUR_DATA_DIR]
huggingface-cli login
hf download --repo-type dataset haodongli/DA-2-Evaluation --local-dir [YOUR_DATA_DIR]
```
2. Unzip the downloaded datasets:
```
tar -zxvf [DATA_NAME].tar.gz
```
3. Set the `datasets_dir` (line 20) in `configs/eval.json` with `YOUR_DATA_DIR`.
4. Run the evaluation command:
```
sh eval.sh
```
5. The results will be saved at `output/eval`.
## 🎓 Citation
If you find our work useful in your research, please consider citing our paper🌹:
```bibtex
@article{li2025depth,
title={DA$^{2}$: Depth Anything in Any Direction},
author={Li, Haodong and Zheng, Wangguangdong and He, Jing and Liu, Yuhao and Lin, Xin and Yang, Xin and Chen, Ying-Cong and Guo, Chunchao},
journal={arXiv preprint arXiv:2509.26618},
year={2025}
}
```
## 🤝 Acknowledgement
This implementation is impossible without the awesome contributions of [MoGe](https://wangrc.site/MoGePage/), [UniK3D](https://lpiccinelli-eth.github.io/pub/unik3d/), [Lotus](https://lotus3d.github.io/), [Marigold](https://marigoldmonodepth.github.io/), [DINOv2](https://github.com/facebookresearch/dinov2), [Accelerate](https://github.com/huggingface/accelerate), [Gradio](https://github.com/gradio-app/gradio), [HuggingFace Hub](https://github.com/huggingface/huggingface_hub), and [PyTorch](https://pytorch.org/) to the open-cource community.
|