# Video-Depth-Anything **Repository Path**: grang/Video-Depth-Anything ## Basic Information - **Project Name**: Video-Depth-Anything - **Description**: No description available - **Primary Language**: Unknown - **License**: Apache-2.0 - **Default Branch**: main - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2026-03-24 - **Last Updated**: 2026-03-24 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README

Video Depth Anything

[**Sili Chen**](https://github.com/SiliChen321) · [**Hengkai Guo**](https://guohengkai.github.io/)† · [**Shengnan Zhu**](https://github.com/Shengnan-Zhu) · [**Feihu Zhang**](https://github.com/zhizunhu)
[**Zilong Huang**](http://speedinghzl.github.io/) · [**Jiashi Feng**](https://scholar.google.com.sg/citations?user=Q8iay0gAAAAJ&hl=en) · [**Bingyi Kang**](https://bingykang.github.io/)†
ByteDance
†Corresponding author Paper PDF Project Page
This work presents **Video Depth Anything** based on [Depth Anything V2](https://github.com/DepthAnything/Depth-Anything-V2), which can be applied to arbitrarily long videos without compromising quality, consistency, or generalization ability. Compared with other diffusion-based models, it enjoys faster inference speed, fewer parameters, and higher consistent depth accuracy. ![teaser](assets/teaser_video_v2.png) ## News - **2025-09-12:** Support streaming mode for metric depth models. - **2025-08-28:** Release ViT-base model for relative depth and ViT-small/base models for video metric depth. - **2025-07-03:** 🚀🚀🚀 Release an experimental version of training-free **streaming video depth estimation**. - **2025-07-03:** Release our implementation of [training loss](https://github.com/DepthAnything/Video-Depth-Anything/tree/main/loss). - **2025-04-25:** 🌟🌟🌟 Release metric depth model based on Video-Depth-Anything-Large. - **2025-04-05:** Our paper has been accepted for a **highlight** presentation at [CVPR 2025](https://cvpr.thecvf.com/) (13.5% of the accepted papers). - **2025-03-11:** Add full dataset inference and evaluation [scripts](https://github.com/DepthAnything/Video-Depth-Anything/tree/main/benchmark). - **2025-02-08:** Enable autocast inference. Support grayscale video, NPZ and EXR output formats. - **2025-01-21:** Paper, project page, code, models, and demo are all released. ## Release Notes - **2025-08-28:** 🚀🚀🚀 Metric depth models released | δ1 | MoGe-2-L | UniDepthV2-L | DepthPro | VDA-S-Metric | VDA-B-Metric | VDA-L-Metric | |:-|:-:|:-:|:-:|:-:|:-:|:-:| | KITTI | 0.415 | **0.982** | 0.822 | 0.877 | 0.887 | *0.910* | | NYUv2 | *0.967* | **0.989** | 0.953 | 0.850| 0.883 | 0.908 | | **TAE** | | | | | | | | Scannet | 2.56 | 1.41 | 2.73 | 1.48 | *1.26* | **1.09** | - **2025-02-08:** 🚀🚀🚀 Inference speed and memory usage improvement
Model Latency (ms) GPU VRAM (GB)
FP32 FP16 FP32 FP16
Video-Depth-Anything-Small 9.1 7.5 7.3 6.8
Video-Depth-Anything-Large 67 14 26.7 23.6
The Latency and GPU VRAM results are obtained on a single A100 GPU with input of shape 1 x 32 x 518 × 518. ## Pre-trained Models We provide **several models** of varying scales for robust and consistent video depth estimation. | Relative Depth Model | Params | Checkpoint | |:-|-:|:-:| | Video-Depth-Anything-Small | 28.4M | [Download](https://huggingface.co/depth-anything/Video-Depth-Anything-Small/resolve/main/video_depth_anything_vits.pth?download=true) | | Video-Depth-Anything-Base | 113.1M | [Download](https://huggingface.co/depth-anything/Video-Depth-Anything-Base/blob/main/video_depth_anything_vitb.pth) | | Video-Depth-Anything-Large | 381.8M | [Download](https://huggingface.co/depth-anything/Video-Depth-Anything-Large/resolve/main/video_depth_anything_vitl.pth?download=true) | | **Metric Depth Model** | **Params** | **Checkpoint** | | Metric-Video-Depth-Anything-Small | 28.4M | [Download](https://huggingface.co/depth-anything/Metric-Video-Depth-Anything-Small/blob/main/metric_video_depth_anything_vits.pth) | | Metric-Video-Depth-Anything-Base | 113.1M | [Download](https://huggingface.co/depth-anything/Metric-Video-Depth-Anything-Base/blob/main/metric_video_depth_anything_vitb.pth) | | Metric-Video-Depth-Anything-Large | 381.8M | [Download](https://huggingface.co/depth-anything/Metric-Video-Depth-Anything-Large/resolve/main/metric_video_depth_anything_vitl.pth) | ## Usage ### Preparation ```bash git clone https://github.com/DepthAnything/Video-Depth-Anything cd Video-Depth-Anything pip install -r requirements.txt ``` Download the checkpoints listed [here](#pre-trained-models) and put them under the `checkpoints` directory. ```bash bash get_weights.sh ``` ### Run inference on a video We support both relative depth and metric depth: ```bash # For relative depth python3 run.py --input_video ./assets/example_videos/davis_rollercoaster.mp4 --output_dir ./outputs --encoder vitl # For metric depth python3 run.py --input_video ./assets/example_videos/davis_rollercoaster.mp4 --output_dir ./outputs --encoder vitl --metric ``` Options: - `--input_video`: path of input video - `--output_dir`: path to save the output results - `--input_size` (optional): By default, we use input size `518` for model inference. - `--max_res` (optional): By default, we use maximum resolution `1280` for model inference. - `--encoder` (optional): `vits` for Video-Depth-Anything-Small, `vitb` for Video-Depth-Anything-Base, `vitl` for Video-Depth-Anything-Large. - `--max_len` (optional): maximum length of the input video, `-1` means no limit - `--target_fps` (optional): target fps of the input video, `-1` means the original fps - `--metric` (optional): use metric depth models trained on Virtual KITTI and IRS datasets - `--fp32` (optional): Use `fp32` precision for inference. By default, we use `fp16`. - `--grayscale` (optional): Save the grayscale depth map, without applying color palette. - `--save_npz` (optional): Save the depth map in `npz` format. - `--save_exr` (optional): Save the depth map in `exr` format. ### Run inference on a video using streaming mode (Experimental features) We implement an experimental streaming mode **without training**. In details, we save the hidden states of temporal attentions for each frames in the caches, and only send a single frame into our video depth model during inference by reusing these past hidden states in temporal attentions. We hack our pipeline to align the original inference setting in the offline mode. Due to the inevitable gap between training and testing, we observe a **performance drop** between the streaming model and the offline model (e.g. the `d1` of ScanNet drops from `0.926` to `0.836`). Finetuning the model in the streaming mode will greatly improve the performance. We leave it for future work. To run the streaming model: ```bash # For relative depth python3 run_streaming.py --input_video ./assets/example_videos/davis_rollercoaster.mp4 --output_dir ./outputs_streaming --encoder vitl # For metric depth python3 run_streaming.py --input_video ./assets/example_videos/davis_rollercoaster.mp4 --output_dir ./outputs_streaming --encoder vitl --metric ``` Options: - `--input_video`: path of input video - `--output_dir`: path to save the output results - `--input_size` (optional): By default, we use input size `518` for model inference. - `--max_res` (optional): By default, we use maximum resolution `1280` for model inference. - `--encoder` (optional): `vits` for Video-Depth-Anything-Small, `vitb` for Video-Depth-Anything-Base, `vitl` for Video-Depth-Anything-Large. - `--max_len` (optional): maximum length of the input video, `-1` means no limit - `--target_fps` (optional): target fps of the input video, `-1` means the original fps - `--metric` (optional): use metric depth models trained on Virtual KITTI and IRS datasets - `--fp32` (optional): Use `fp32` precision for inference. By default, we use `fp16`. - `--grayscale` (optional): Save the grayscale depth map, without applying color palette. ## Training Loss Our training loss is in `loss/` directory. Please see the `loss/test_loss.py` for usage. ## Benchmark Please refer to [Benchmark](./benchmark/README.md). ## Citation If you find this project useful, please consider citing: ```bibtex @article{video_depth_anything, title={Video Depth Anything: Consistent Depth Estimation for Super-Long Videos}, author={Chen, Sili and Guo, Hengkai and Zhu, Shengnan and Zhang, Feihu and Huang, Zilong and Feng, Jiashi and Kang, Bingyi}, journal={arXiv:2501.12375}, year={2025} } ``` ## LICENSE Video-Depth-Anything-Small model is under the Apache-2.0 license. Video-Depth-Anything-Base/Large model is under the CC-BY-NC-4.0 license. For business cooperation, please send an email to Hengkai Guo at guohengkaighk@gmail.com.