{
  "architectures": [
    "DepthAnything3ForDepthEstimation"
  ],
  "auto_map": {
    "AutoConfig": "configuration_depth_anything_3.DepthAnything3Config",
    "AutoModel": "modeling_depth_anything_3.DepthAnything3ForDepthEstimation",
    "AutoModelForDepthEstimation": "modeling_depth_anything_3.DepthAnything3ForDepthEstimation"
  },
  "da3_config": {
    "__object__": {
      "args": "as_params",
      "name": "DepthAnything3Net",
      "path": "depth_anything_3.model.da3"
    },
    "cam_dec": {
      "__object__": {
        "args": "as_params",
        "name": "CameraDec",
        "path": "depth_anything_3.model.cam_dec"
      },
      "dim_in": 1536
    },
    "cam_enc": {
      "__object__": {
        "args": "as_params",
        "name": "CameraEnc",
        "path": "depth_anything_3.model.cam_enc"
      },
      "dim_out": 768
    },
    "head": {
      "__object__": {
        "args": "as_params",
        "name": "DualDPT",
        "path": "depth_anything_3.model.dualdpt"
      },
      "dim_in": 1536,
      "features": 128,
      "out_channels": [
        96,
        192,
        384,
        768
      ],
      "output_dim": 2
    },
    "net": {
      "__object__": {
        "args": "as_params",
        "name": "DinoV2",
        "path": "depth_anything_3.model.dinov2.dinov2"
      },
      "alt_start": 4,
      "cat_token": true,
      "name": "vitb",
      "out_layers": [
        5,
        7,
        9,
        11
      ],
      "qknorm_start": 4,
      "rope_start": 4
    }
  },
  "dtype": "float32",
  "model_name": "da3-base",
  "model_type": "depth_anything_3",
  "transformers_version": "5.20.2"
}