depth-estimation-ros

Support Backend

• TensorRT

Support Model

Mono Depth

• Depth-Anything-V2
  • Reference of model converting to TensorRT Engine
    • https://github.com/spacewalk01/depth-anything-tensorrt

Stereo

• CREStereo
  • ONNX model download
    • https://github.com/PINTO0309/PINTO_model_zoo/tree/main/284_CREStereo
• Fast-ACVNet
  • ONNX model download
    • https://github.com/PINTO0309/PINTO_model_zoo/tree/main/338_Fast-ACVNet

Topic

Subscribe

Mono Depth

• image_raw (sensor_msgs::msg::Image)

Stereo Depth

• /camera/camera/infra1/image_rect_raw (sensor_msgs::msg::Image)
• /camera/camera/infra2/image_rect_raw (sensor_msgs::msg::Image)
• /camera/camera/infra1/camera_info (sensor_msgs::msg::CameraInfo)
• /camera/camera/infra2/camera_info (sensor_msgs::msg::CameraInfo)

Publish

• depth_estimation_ros/depth/image_raw (sensor_msgs::msg::Image)

  • Depth image
  • format: mono16

• depth_estimation_ros/color/image_raw (sensor_msgs::msg::Image)

  • Colorized depth image.
  • format: bgr8
  • colormap: jet

Parameters

• model_path (str)
  • model absolute or relative path.
• backend (str)
  • default: "tensorrt"
  • inference backend.
• model_type (str)
  • default: "mono"
  • specify "mono" or "stereo" model.
• model_input_normalize (bool)
  • default: true
  • Whether to normalize model input value with mean and std.
• model_input_mean (vector of double)
  • default: {0.485, 0.456, 0.406}
• model_input_std (vector of double)
  • default: {0.229, 0.224, 0.225}
• model_swap_r_b (bool)
  • default: false
  • Whether to swap model input Red channel and Blue channel.
• stereo_baseline_meter (double)
  • default: 0.05
  • Specify distance between left and right camera.
• depth_scale (double)
  • default: 0.001
  • Scale value when converted to uint16_t image.
• depth_offset (double)
  • default: 0.0
• max_depth_meter (double)
  • default: 20.0
  • Replace distance values greater than this threshold with zeros.
• min_depth_meter (double)
  • default: 0.0
  • Replace distance values smaller than this threshold with zeros
• publish_colored_depth_image (bool)
  • default: true
  • Whether to publish colored_depth_image.
• publish_depth_image (bool)
  • default: true
  • Whether to publish depth_image.

• imshow (bool)
  • default: true
  • Whether to show colored depth image.
• tensorrt_device (int)
  • default: 0
  • CUDA device ID used by TensorRT.