As this model also supports image guided detections, I don't see an implementation from autodistill for the same. Are there any ways I can make it work while being in Autodistill env?