YOLOv10 speed slower than v8

[rohansaw]

In the Pre-Trained Models Object Detection Overview, the FPS for YOLOv8n with a batchsize of 8 are 465FPS, while for v10n it is only 286FPS. This seems odd, as YOLOv10 is supposed to be quite a bit faster than v8 (https://docs.ultralytics.com/models/yolov10/#accuracy-enhancements).

Is this being investigated or is there a clear reason for this? I would have expected v10 to benefit even more from acceleration of the AI accelerator as it does not rely on NMS.