README.md
| 1 | --- |
| 2 | license: cc-by-nc-4.0 |
| 3 | |
| 4 | language: |
| 5 | - en |
| 6 | pipeline_tag: depth-estimation |
| 7 | library_name: depth-anything-v2 |
| 8 | tags: |
| 9 | - depth |
| 10 | - relative depth |
| 11 | --- |
| 12 | |
| 13 | # Depth-Anything-V2-Large |
| 14 | |
| 15 | ## Introduction |
| 16 | Depth Anything V2 is trained from 595K synthetic labeled images and 62M+ real unlabeled images, providing the most capable monocular depth estimation (MDE) model with the following features: |
| 17 | - more fine-grained details than Depth Anything V1 |
| 18 | - more robust than Depth Anything V1 and SD-based models (e.g., Marigold, Geowizard) |
| 19 | - more efficient (10x faster) and more lightweight than SD-based models |
| 20 | - impressive fine-tuned performance with our pre-trained models |
| 21 | |
| 22 | ## Installation |
| 23 | |
| 24 | ```bash |
| 25 | git clone https://huggingface.co/spaces/depth-anything/Depth-Anything-V2 |
| 26 | cd Depth-Anything-V2 |
| 27 | pip install -r requirements.txt |
| 28 | ``` |
| 29 | |
| 30 | ## Usage |
| 31 | |
| 32 | Download the [model](https://huggingface.co/depth-anything/Depth-Anything-V2-Large/resolve/main/depth_anything_v2_vitl.pth?download=true) first and put it under the `checkpoints` directory. |
| 33 | |
| 34 | ```python |
| 35 | import cv2 |
| 36 | import torch |
| 37 | |
| 38 | from depth_anything_v2.dpt import DepthAnythingV2 |
| 39 | |
| 40 | model = DepthAnythingV2(encoder='vitl', features=256, out_channels=[256, 512, 1024, 1024]) |
| 41 | model.load_state_dict(torch.load('checkpoints/depth_anything_v2_vitl.pth', map_location='cpu')) |
| 42 | model.eval() |
| 43 | |
| 44 | raw_img = cv2.imread('your/image/path') |
| 45 | depth = model.infer_image(raw_img) # HxW raw depth map |
| 46 | ``` |
| 47 | |
| 48 | ## Citation |
| 49 | |
| 50 | If you find this project useful, please consider citing: |
| 51 | |
| 52 | ```bibtex |
| 53 | @article{depth_anything_v2, |
| 54 | title={Depth Anything V2}, |
| 55 | author={Yang, Lihe and Kang, Bingyi and Huang, Zilong and Zhao, Zhen and Xu, Xiaogang and Feng, Jiashi and Zhao, Hengshuang}, |
| 56 | journal={arXiv:2406.09414}, |
| 57 | year={2024} |
| 58 | } |
| 59 | |
| 60 | @inproceedings{depth_anything_v1, |
| 61 | title={Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data}, |
| 62 | author={Yang, Lihe and Kang, Bingyi and Huang, Zilong and Xu, Xiaogang and Feng, Jiashi and Zhao, Hengshuang}, |
| 63 | booktitle={CVPR}, |
| 64 | year={2024} |
| 65 | } |