Desktop2Stereo: a 2D Desktop to 3D Stereo App

中文版本 FPS Performance

A universal real-time 2D to 3D App that supports AMD/NVIDIA/Intel/Qualcomm GPU/Apple Silicon devices on Windows/Mac/Linux OS, powered by Depth Estimation AI Models

Alternative Download Link

Quark NetDrive
Access code: 1vcn

Supported Hardware

1. AMD GPU
2. NVIDIA GPU
3. Apple Silicon Chip (M1, M2, M3, M4, ...)
4. DirectML compatible devices (Intel Arc/Iris GPU, Qualcomm® Adreno GPU, etc. Windows only)

Supported OS

1. Windows 10/11 (x64/Arm64)
2. MacOS 10.16 or later
3. Linux OS (beta)

Install and Run

Windows

1. Install latest GPU driver
   AMD GPU: Download latest GPU driver from AMD Drivers and Support for Processors and Graphics. NVIDIA GPU: Download latest GPU driver from NVIDA Official GeForce Drivers.
   Intel GPU: Download latest GPU driver from Download Intel Drivers and Software.
   Qualcomm GPU: Download latest GPU driver from Qualcomm® Adreno™ Windows Graphics Drivers for Snapdragon® X Platform.
   Other DirectML devices: Please install the latest hardware driver accordingly.
2. Install Python 3.10
   Download from Python.org and install.
3. Download Desktop2Stereo app
   Download the Desktop2Stereo.zip and unzip it to local disk.
4. Install python environment
   AMD/Intel/Qualcomm GPU and other DirectML compatible devies: Doulbe click install-dml.bat.
   NVIDIA GPU: Doulbe click install-cuda.bat.
5. Run Stereo2Desktop GUI application
   Doulbe click run.bat.

MacOS

1. Install Python 3.10
   Download from Python.org and install.
2. Download Desktop2Stereo app
   Download the Desktop2Stereo.zip and unzip it to local disk.
3. Install Python environment
   Doulbe click install-mps executable. (Please allow open in Privacy and Security Settings)
4. Run Stereo2Desktop GUI application
   Doulbe click run_mac executable. (Please allow open in Privacy and Security Settings, "Screen Recording" permission is required)

Linux (Beta)

1. Install latest GPU driver
   AMD GPU: Download latest GPU driver and ROCm from AMD Drivers and Support for Processors and Graphics. NVIDIA GPU: Download latest GPU driver from AMD Drivers and Support for Processors and Graphics.
2. Install Python 3.10

   # Example: Ubuntu
   sudo add-apt-repository ppa:savoury1/python
   sudo apt update
   sudo apt-get install python3.10

3. Download Desktop2Stereo app
   Download the Desktop2Stereo.zip and unzip it to local disk.
4. Install Python environment
   AMD GPU: Run install-rocm.bash:

   bash install-rocm.bash

   NVIDIA GPU: Run install-cuda.bash:

   bash install-cuda.bash

5. Run Stereo2Desktop GUI application
   Run run_linux.bash:

   bash run_linux.bash

Desktop2Stereo GUI App

Quick Run Desktop2Stereo

Just use the default settnigs and click Run, and then click OK to run the Stereo Viewer window.

Stereo Viewer Window

1. Use ← Left or → Right arrow keys to switch the Stereo Viewer window to second (virtual) monitor display.
2. Set your video/game on the main screen (full screen mode if you needed).
3. Click the Stereo Viewer on second (virtual) monitor display to make sure the Stereo Viewer is the 1st active application. Press space to toggle full screen mode.
4. Now you can use AR/VR to view the SBS or TAB output.

• AR need to switch to 3D mode to connect as a 3840*1080 (Full Side-by-Side, FUll-SBS) display.
  
• VR need to use 2nd Display/Virtual Display (VDD) with Desktop+[PC VR] or Virtual Desktop[PC/Standalone VR] or OBS + Wolvic Browser [Standalone VR] to comopose the SBS (Side-by-Side) / TAB (Top-and-Bottom) display to 3D.
• You can use Tab key to toggle SBS/TAB mode.
  
  

5. Real-time modification of depth strength.
   Use ↑ Up or ↓ Right arrow keys to increase/decrease the depth strength by a step of 0.1. To reset press 0 key. The defination of depth strength is in the detailed settings session.
6. Press Esc to exit the Stereo Viewer.

Detailed Settings

All optional settings can be modified on the GUI window and saved to the settings.yaml. Each time you click Run, the settings will be saved automatically, and clicking Reset will restore the default settings.

1. Set Language
   English (EN) and Simplified Chinese (CN) are supported.
2. Monitor index
   Default is your Primary Monitor (mostly shall follow the monitor numbers in your system settings).
3. Device
   Default shall be your GPU (CUDA/DirectML/MPS), or CPU if you don't have a compatible GPU device.
4. FP16 Recommanded for most computing devices for better performance. If your device does not support FP16 DataType, disable it.
5. Show FPS Show FPS on the topbar of the Stereo Viewer.
6. Output Resolution Default output resolution is 1080 (1080p, 1920x1080) for a smoother experience. 2160 (4K, 3840x2160) and 1440 (2K, 2560x1440) resolutions are also available if you have powerful devices.
7. FPS (frames per second) FPS can set as your monitor refresh rate, default input FPS is 60. It determins the freqency of the screen caputre process (higher FPS does not ensure smoother output, depending on your devices).
8. Depth Resolution
   Higher depth resolution can give better depth details but cause higher GPU usage, which is also related to the model training settings.
   Default depth resolution is set to 384 for balanced performance.
9. Depth Strength
   With higher depth strength, 3D depth effect of the object would be stronger. However, higher value can induce visible artifacts. Default is set to 1.0. The recomanded depth strength range is (1, 5).
10. IPD (Interpupillary Distance)
    IPD is the distance between the centers of your pupils, it affects how your brain interprets stereoscopic 3D. The default IPD is 0.064 in meter (m), which is the average human IPD value.
11. Download Path
    Default download path is the models folder under the working directory.
12. Depth Model Modify the depth model id from HuggingFace, the model id under depth_model mostly shall ends with -hf.
    Large model can cause higher GPU usage and latency.
    Default depth model: depth-anything/Depth-Anything-V2-Small-hf
    Currently supported models:

    Model List:
    - depth-anything/Depth-Anything-V2-Large-hf
    - depth-anything/Depth-Anything-V2-Base-hf
    - depth-anything/Depth-Anything-V2-Small-hf
    - depth-anything/Depth-Anything-V2-Metric-Outdoor-Large-hf
    - depth-anything/Depth-Anything-V2-Metric-Outdoor-Base-hf
    - depth-anything/Depth-Anything-V2-Metric-Outdoor-Small-hf
    - depth-anything/Depth-Anything-V2-Metric-Indoor-Large-hf
    - depth-anything/Depth-Anything-V2-Metric-Indoor-Base-hf
    - depth-anything/Depth-Anything-V2-Metric-Indoor-Small-hf
    - LiheYoung/depth-anything-large-hf
    - LiheYoung/depth-anything-base-hf
    - LiheYoung/depth-anything-small-hf
    - xingyang1/Distill-Any-Depth-Large-hf
    - xingyang1/Distill-Any-Depth-Small-hf
    - apple/DepthPro-hf # Depth: 1536
    - Intel/dpt-large # Slow, NOT recommand

You can also manually add the hugging face models in the settings.yaml which including the following:
model.safetensors
config.json
preprocessor_config.json
13. HF Endpoint (Hugging Face)
HF-Mirror is a mirror site of the original Hugging Face site hosting AI models. The depth model will automatically be downloaded to Download Path from Hugging Face at the first run.

References

@article{depth_anything_v2,
  title={Depth Anything V2},
  author={Yang, Lihe and Kang, Bingyi and Huang, Zilong and Zhao, Zhen and Xu, Xiaogang and Feng, Jiashi and Zhao, Hengshuang},
  journal={arXiv:2406.09414},
  year={2024}
}

@inproceedings{depth_anything_v1,
  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}
}

@article{he2025distill,
  title   = {Distill Any Depth: Distillation Creates a Stronger Monocular Depth Estimator},
  author  = {Xiankang He and Dongyan Guo and Hongji Li and Ruibo Li and Ying Cui and Chi Zhang},
  year    = {2025},
  journal = {arXiv preprint arXiv: 2502.19204}
}

@inproceedings{Bochkovskii2024:arxiv,
  author     = {Aleksei Bochkovskii and Ama\"{e}l Delaunoy and Hugo Germain and Marcel Santos and
               Yichao Zhou and Stephan R. Richter and Vladlen Koltun},
  title      = {Depth Pro: Sharp Monocular Metric Depth in Less Than a Second},
  booktitle  = {International Conference on Learning Representations},
  year       = {2025},
  url        = {https://arxiv.org/abs/2410.02073},
}

@article{DBLP:journals/corr/abs-2103-13413,
  author    = {Ren{\'{e}} Ranftl and
               Alexey Bochkovskiy and
               Vladlen Koltun},
  title     = {Vision Transformers for Dense Prediction},
  journal   = {CoRR},
  volume    = {abs/2103.13413},
  year      = {2021},
  url       = {https://arxiv.org/abs/2103.13413},
  eprinttype = {arXiv},
  eprint    = {2103.13413},
  timestamp = {Wed, 07 Apr 2021 15:31:46 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2103-13413.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}