> ## Documentation Index
> Fetch the complete documentation index at: https://docs.generalrobotics.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# DepthAnything v2 

The `DepthAnything_V2` class implements a wrapper for the DepthAnything\_V2 model, which estimates depth
maps from RGB images. The model supports 'metric' and 'relative' modes, which load
different pre-trained models based on the specified mode. We use the VIT Large encoder.

<ResponseField name="class DepthAnything_V2()">
  <ResponseField name="mode" type="string" default="metric">
    Flag to specify the mode of the model. Can be 'metric' or 'relative'. Defaults to 'metric'.
  </ResponseField>

  <ResponseField name="use_local" type="boolean" default="False">
    If True, inference call is run on the local VM, else offloaded onto GRID-Cortex. Defaults to False.
  </ResponseField>
</ResponseField>

<ResponseField name="def run()">
  <ResponseField name="rgbimage" type="np.ndarray" required="True">
    The input RGB image of shape $(M,N,3)$.
  </ResponseField>

  <ResponseField name="Returns" type="np.ndarray">
    The predicted depth map of shape $(M,N)$.
  </ResponseField>
</ResponseField>

<RequestExample>
  ```python Inference Call theme={null}
  from grid.model.perception.depth.depth_anything_v2 import DepthAnything_V2
  car = AirGenCar()

  # We will be capturing an image from the AirGen simulator 
  # and run model inference on it.

  img =  car.getImage("front_center", "rgb").data

  model = DepthAnything_V2(use_local = False, mode='relative')
  result = model.run(rgbimage=img)
  print(result.shape)
  ```
</RequestExample>

<Tabs>
  <Tab title="License">
    This code is licensed under the Apache 2.0 License.
  </Tab>

  <Tab title="Source">
    [https://huggingface.co/spaces/depth-anything/Depth-Anything-V2](https://huggingface.co/spaces/depth-anything/Depth-Anything-V2)
  </Tab>
</Tabs>