It's XYXY top left,bottom right. I guess the negative values are just the model predicting that the object extends outside the image. Leaving this open because it's still a problem that this isn't documented well.

Is it xyxy? xywh? If xywh, is the xy the center or top-left? I ran a yolo pipeline on an image and there are negative numbers in the bounding box ouput. Are the coordinates relative to the image center or something?

No info to be found in any of these places that would make sense

• YOLOOutput class source code
  
• YOLOPipeline class source code
• user guide
• object detection use case guide (https://github.com/neuralmagic/deepsparse/blob/main/docs/use-cases/cv/object-detection-yolov5.md)

I would be more than happy to get a "RTFM" response if I could just get a link to where this is documented.