Can we change the default values of iou and conf for the pre-trained models during the inference?

[halqadasi]

I have used the pre-trained model below from the PyTorch hub in the inference time. The output of the model here includes the NMS, my question is: Can we change the default values of iou and conf which are 0.45 and 0.25, respectively during the inference?

model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True)

[glenn-jocher]

@halqadasi yes, you can adjust the iou and conf thresholds during inference with YOLOv8. When you load the model, you can set the conf threshold for object confidence and the iou threshold for non-maximum suppression directly on the model object. For example, after loading the model, you can modify these attributes like so:

model.conf = 0.30  # Set a new confidence threshold
model.iou = 0.50   # Set a new IoU threshold for NMS

Just replace 0.30 and 0.50 with your desired confidence and IoU thresholds, respectively. This will affect how the model filters predictions during the inference process. 🚀

For more detailed information on how to use and configure YOLOv8 models, please refer to our documentation at https://docs.ultralytics.com.