Devesh Kumar comments

Results 5 comments of


                                            Devesh Kumar

Quantization-Aware Training for YOLOv8

> @hamedgorji hi there! 👋 > > Thanks so much for reaching out and for your kind words about YOLOv8. We truly appreciate your suggestion regarding quantization-aware training. It’s feedback...

Quantization-Aware Training for YOLOv8

QAT is really important if ultralytics want to penitrate the EDGE computing market. I really need this functionality and want to help in PR feature enhancement.

any plan to release a pre-trained model on DOTAV2 or xView satellite imagery dataset?

that great. can you please provide YOLO-World satellite data pre trained model so that I can see a demo. then I can train on my dataset.

inference No module named 'detectron2'

Even after installing detectron2 , i am geeeting this error.. please help.. /content/drive/MyDrive/htcr_sam_road/sam_road Traceback (most recent call last): File "/content/drive/MyDrive/htcr_sam_road/sam_road/inferencer.py", line 247, in net = SAMRoad(config) File "/content/drive/MyDrive/htcr_sam_road/sam_road/model.py", line 363,...

export yolo 11 in ONXX with options doesn't change the accuracy and confidence score. are these option working in yolov11?

@glenn-jocher @Y-T-G During quantization of YOLOv11, the confidence score for the same object should demonstrably decrease slightly. However, I did not observe this drop. In both Method 1 (using half=True...