{
  "major_version": 0,
  "minor_version": 1,
  "model_name": "vad-carla-tiny",
  "training_dataset": "Bench2Drive CARLA",
  "release_date": "2025-11-04",
  "description": "VAD model trained on Bench2Drive CARLA dataset with 6-camera surround view",

  "image_normalization": {
    "mean": [123.675, 116.28, 103.53],
    "std": [58.395, 57.12, 57.375],
    "format": "RGB",
    "note": "ImageNet normalization statistics"
  },

  "input_specs": {
    "target_image_width": 640,
    "target_image_height": 384,
    "num_cameras": 6,
    "camera_order": ["FRONT", "BACK", "FRONT_LEFT", "BACK_LEFT", "FRONT_RIGHT", "BACK_RIGHT"]
  },

  "class_definitions": {
    "map_classes": ["Broken", "Solid", "SolidSolid", "Center", "TrafficLight", "StopSign"],
    "object_classes": ["car", "van", "truck", "bicycle", "traffic_sign", "traffic_cone", "traffic_light", "pedestrian", "others"]
  },

  "network_architecture": {
    "bev": {
      "height": 106,
      "width": 200,
      "feature_dim": 256,
      "downsample_factor": 32
    },
    "transformer": {
      "num_decoder_layers": 3
    },
    "prediction": {
      "num_queries": 300,
      "num_classes": 9,
      "bbox_pred_dim": 10,
      "trajectory_modes": 6,
      "timesteps": 6,
      "bbox_format": "[cx, cy, w, l, cz, h, sin, cos, vx, vy]"
    },
    "planning": {
      "ego_commands": 6,
      "timesteps": 6
    },
    "map": {
      "num_queries": 100,
      "num_classes": 6,
      "points_per_polyline": 20
    },
    "vehicle_state": {
      "can_bus_dim": 18
    }
  },

  "model_files": {
    "backbone": "vad-carla-tiny_backbone.onnx",
    "head": "vad-carla-tiny_head.onnx",
    "head_no_prev": "vad-carla-tiny_head_no_prev.onnx"
  },

  "compatibility": {
    "autoware_tensorrt_vad_version": ">= 0.1.0",
    "tensorrt_version": ">= 8.6",
    "cuda_version": ">= 11.4"
  }
}