Depth Anything V2 Project

This project utilizes the Depth Anything V2 model (safetensors formats) Original repository.

Installation

Follow these steps to set up the project environment and install dependencies.

1. Clone the Repository (Optional)

If you haven't already, clone the repository to your local machine:

git clone https://github.com/yushan777/depth-anything-v2.git
cd depth-anything-v2

2. Create a Python Virtual Environment

It's highly recommended to use a virtual environment to manage project dependencies.

python3 -m venv venv 
# Or python -m venv venv

• This command creates a directory named venv in your project folder containing a dedicated Python installation.

3. Activate the Virtual Environment

You need to activate the environment before installing packages. The command differs based on your operating system:

• macOS / Linux (bash/zsh):

  source venv/bin/activate

• Windows (Command Prompt):

  venv\Scripts\activate.bat

• Windows (PowerShell):

  venv\Scripts\Activate.ps1

  (Note: You might need to adjust your PowerShell execution policy: Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope Process)

  You'll know the environment is active when you see (venv) prepended to your command prompt.

4. Install PyTorch (Hardware-Specific)

Run the provided script to detect your hardware and install the correct PyTorch build (including torchvision and torchaudio):

python install_torch.py

• This script checks for NVIDIA GPUs (installing the CUDA version if found) or Apple Silicon (installing the standard version with MPS support).

5. Install Remaining Dependencies

Install the other required packages listed in requirements.txt:

pip install -r requirements.txt

Usage Example

To run the model on an image:

python main.py \
    --input path/to/your/image.jpg \
    ----output-dir path/to/save/depth_maps \
    --model depth_anything_v2_vitl_fp32.safetensors

• Replace path/to/your/image.jpg and path/to/save/depth_maps with your actual file and dir paths.
• Choose the desired model using the --model argument.
• The script will automatically attempt to use a GPU (CUDA or MPS) if available, falling back to CPU otherwise.

6. Deactivate the Virtual Environment (When Done)