Diff between camera_obstacle_detection and fusion_camera_detection ?

[lovelydett]

I noticed both of them process Image and output obstacles. In my understanding, camera_detection runs some DNN tasks (e.g., Yolo) to perform one-stage recognition for obstacles in one Image frame, but fusion_camera_detection seemingly does similar things. Specifically, the word "fusion" confuses me. I wonder if the fusion here indicates the fusion between multiple cameras (unlike the "fusion" in fusion_component which fuses multiple sensors)

Thank you for your time!

[daohu527]

@lovelydett Yes they are almost the same.

• fusion_camera_detection_component. Older version for camera 3d object detection
• camera_obstacle_detection_component. Newer versions for camera 3d object detection use SMOKE

The biggest difference between them is the camera 2d to 3d conversion. camera_obstacle_detection_component outputs 2d and 3d directly through the model.

The details may be slightly different, but we will update in the near future such as merging them.