Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data

Lihe Yang¹ · Bingyi Kang²⁺ · Zilong Huang² · Xiaogang Xu^3,4 · Jiashi Feng² · Hengshuang Zhao¹⁺

¹The University of Hong Kong · ²TikTok · ³Zhejiang Lab · ⁴Zhejiang University

⁺corresponding authors

CVPR 2024

This work presents Depth Anything, a highly practical solution for robust monocular depth estimation by training on a combination of 1.5M labeled images and 62M+ unlabeled images.

News

• 2024-02-27: Depth Anything is accepted by CVPR 2024.
• 2024-02-05: Depth Anything Gallery is released. Thank all the users!
• 2024-02-02: Depth Anything serves as the default depth processor for InstantID and InvokeAI.
• 2024-01-25: Support video depth visualization. An online demo for video is also available.
• 2024-01-23: The new ControlNet based on Depth Anything is integrated into ControlNet WebUI and ComfyUI's ControlNet.
• 2024-01-23: Depth Anything ONNX and TensorRT versions are supported.
• 2024-01-22: Paper, project page, code, models, and demo (HuggingFace, OpenXLab) are released.

Features of Depth Anything

If you need other features, please first check existing community supports.

• Relative depth estimation:

  Our foundation models listed here can provide relative depth estimation for any given image robustly. Please refer here for details.

• Metric depth estimation

  We fine-tune our Depth Anything model with metric depth information from NYUv2 or KITTI. It offers strong capabilities of both in-domain and zero-shot metric depth estimation. Please refer here for details.

• Better depth-conditioned ControlNet

  We re-train a better depth-conditioned ControlNet based on Depth Anything. It offers more precise synthesis than the previous MiDaS-based ControlNet. Please refer here for details. You can also use our new ControlNet based on Depth Anything in ControlNet WebUI or ComfyUI's ControlNet.

• Downstream high-level scene understanding

  The Depth Anything encoder can be fine-tuned to downstream high-level perception tasks, e.g., semantic segmentation, 86.2 mIoU on Cityscapes and 59.4 mIoU on ADE20K. Please refer here for details.

Performance

Here we compare our Depth Anything with the previously best MiDaS v3.1 BEiT_L-512 model.

Please note that the latest MiDaS is also trained on KITTI and NYUv2, while we do not.

Method	Params	KITTI		NYUv2		Sintel		DDAD		ETH3D		DIODE
		AbsRel	$\delta_1$	AbsRel	$\delta_1$	AbsRel	$\delta_1$	AbsRel	$\delta_1$	AbsRel	$\delta_1$	AbsRel	$\delta_1$
MiDaS	345.0M	0.127	0.850	0.048	0.980	0.587	0.699	0.251	0.766	0.139	0.867	0.075	0.942
Ours-S	24.8M	0.080	0.936	0.053	0.972	0.464	0.739	0.247	0.768	0.127	0.885	0.076	0.939
Ours-B	97.5M	0.080	0.939	0.046	0.979	0.432	0.756	0.232	0.786	0.126	0.884	0.069	0.946
Ours-L	335.3M	0.076	0.947	0.043	0.981	0.458	0.760	0.230	0.789	0.127	0.882	0.066	0.952

We highlight the best and second best results in bold and italic respectively (better results: AbsRel $\downarrow$ , $\delta_1 \uparrow$).

Pre-trained models

We provide three models of varying scales for robust relative depth estimation:

Model	Params	Inference Time on V100 (ms)	A100	RTX4090 (TensorRT)
Depth-Anything-Small	24.8M	12	8	3
Depth-Anything-Base	97.5M	13	9	6
Depth-Anything-Large	335.3M	20	13	12

Note that the V100 and A100 inference time (without TensorRT) is computed by excluding the pre-processing and post-processing stages, whereas the last column RTX4090 (with TensorRT) is computed by including these two stages (please refer to Depth-Anything-TensorRT).

You can easily load our pre-trained models by:

from depth_anything.dpt import DepthAnything

encoder = 'vits' # can also be 'vitb' or 'vitl'
depth_anything = DepthAnything.from_pretrained('LiheYoung/depth_anything_{:}14'.format(encoder))

Depth Anything is also supported in transformers. You can use it for depth prediction within 3 lines of code (credit to @niels).

No network connection, cannot load these models?

Click here for solutions

• First, manually download the three checkpoints: depth-anything-large, depth-anything-base, and depth-anything-small.

• Second, upload the folder containing the checkpoints to your remote server.

• Lastly, load the model locally:

from depth_anything.dpt import DepthAnything

model_configs = {
    'vitl': {'encoder': 'vitl', 'features': 256, 'out_channels': [256, 512, 1024, 1024]},
    'vitb': {'encoder': 'vitb', 'features': 128, 'out_channels': [96, 192, 384, 768]},
    'vits': {'encoder': 'vits', 'features': 64, 'out_channels': [48, 96, 192, 384]}
}

encoder = 'vitl' # or 'vitb', 'vits'
depth_anything = DepthAnything(model_configs[encoder])
depth_anything.load_state_dict(torch.load(f'./checkpoints/depth_anything_{encoder}14.pth'))

Note that in this locally loading manner, you also do not have to install the huggingface_hub package. In this way, please feel free to delete this line and the PyTorchModelHubMixin in this line.

Usage

Installation

git clone https://github.com/LiheYoung/Depth-Anything
cd Depth-Anything
pip install -r requirements.txt

Running

python run.py --encoder <vits | vitb | vitl> --img-path <img-directory | single-img | txt-file> --outdir <outdir> [--pred-only] [--grayscale]

Arguments:

• --img-path: you can either 1) point it to an image directory storing all interested images, 2) point it to a single image, or 3) point it to a text file storing all image paths.
• --pred-only is set to save the predicted depth map only. Without it, by default, we visualize both image and its depth map side by side.
• --grayscale is set to save the grayscale depth map. Without it, by default, we apply a color palette to the depth map.

For example:

python run.py --encoder vitl --img-path assets/examples --outdir depth_vis

If you want to use Depth Anything on videos:

python run_video.py --encoder vitl --video-path assets/examples_video --outdir video_depth_vis

Gradio demo

To use our gradio demo locally:

python app.py

You can also try our online demo.

Import Depth Anything to your project

If you want to use Depth Anything in your own project, you can simply follow run.py to load our models and define data pre-processing.

Code snippet (note the difference between our data pre-processing and that of MiDaS)

from depth_anything.dpt import DepthAnything
from depth_anything.util.transform import Resize, NormalizeImage, PrepareForNet

import cv2
import torch
from torchvision.transforms import Compose

encoder = 'vits' # can also be 'vitb' or 'vitl'
depth_anything = DepthAnything.from_pretrained('LiheYoung/depth_anything_{:}14'.format(encoder)).eval()

transform = Compose([
    Resize(
        width=518,
        height=518,
        resize_target=False,
        keep_aspect_ratio=True,
        ensure_multiple_of=14,
        resize_method='lower_bound',
        image_interpolation_method=cv2.INTER_CUBIC,
    ),
    NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
    PrepareForNet(),
])

image = cv2.cvtColor(cv2.imread('your image path'), cv2.COLOR_BGR2RGB) / 255.0
image = transform({'image': image})['image']
image = torch.from_numpy(image).unsqueeze(0)

# depth shape: 1xHxW
depth = depth_anything(image)

Do not want to define image pre-processing or download model definition files?

Easily use Depth Anything through transformers within 3 lines of code! Please refer to these instructions (credit to @niels).

Note: If you encounter KeyError: 'depth_anything', please install the latest transformers from source:

pip install git+https://github.com/huggingface/transformers.git

Click here for a brief demo:

from transformers import pipeline
from PIL import Image

image = Image.open('Your-image-path')
pipe = pipeline(task="depth-estimation", model="LiheYoung/depth-anything-small-hf")
depth = pipe(image)["depth"]

Community Support

We sincerely appreciate all the extensions built on our Depth Anything from the community. Thank you a lot!

Here we list the extensions we have found:

• Depth Anything TensorRT:
  • https://github.com/spacewalk01/depth-anything-tensorrt
  • https://github.com/thinvy/DepthAnythingTensorrtDeploy
  • https://github.com/daniel89710/trt-depth-anything
• Depth Anything ONNX: https://github.com/fabio-sim/Depth-Anything-ONNX
• Depth Anything in Transformers.js (3D visualization): https://huggingface.co/spaces/Xenova/depth-anything-web
• Depth Anything for video (online demo): https://huggingface.co/spaces/JohanDL/Depth-Anything-Video
• Depth Anything in ControlNet WebUI: https://github.com/Mikubill/sd-webui-controlnet
• Depth Anything in ComfyUI's ControlNet: https://github.com/Fannovel16/comfyui_controlnet_aux
• Depth Anything in X-AnyLabeling: https://github.com/CVHub520/X-AnyLabeling
• Depth Anything in OpenXLab: https://openxlab.org.cn/apps/detail/yyfan/depth_anything
• Depth Anything in OpenVINO: https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/280-depth-anything
• Depth Anything ROS:
  • https://github.com/scepter914/DepthAnything-ROS
  • https://github.com/polatztrk/depth_anything_ros
• Depth Anything Android:
  • https://github.com/FeiGeChuanShu/ncnn-android-depth_anything
  • https://github.com/shubham0204/Depth-Anything-Android
• Depth Anything in TouchDesigner: https://github.com/olegchomp/TDDepthAnything
• LearnOpenCV research article on Depth Anything: https://learnopencv.com/depth-anything

If you have your amazing projects supporting or improving (e.g., speed) Depth Anything, please feel free to drop an issue. We will add them here.

Acknowledgement

We would like to express our deepest gratitude to AK(@_akhaliq) and the awesome HuggingFace team (@niels, @hysts, and @yuvraj) for helping improve the online demo and build the HF models.

Besides, we thank the MagicEdit team for providing some video examples for video depth estimation, and Tiancheng Shen for evaluating the depth maps with MagicEdit.

Citation

If you find this project useful, please consider citing:

@inproceedings{depthanything,
      title={Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data}, 
      author={Yang, Lihe and Kang, Bingyi and Huang, Zilong and Xu, Xiaogang and Feng, Jiashi and Zhao, Hengshuang},
      booktitle={CVPR},
      year={2024}
}