search

Alan-D-Chen / CDIoU-CDIoUloss

🔥CDIoU and CDIoU loss is like a convenient plug-in that can be used in multiple models. CDIoU and CDIoU loss have different excellent performances in several models such as Faster R-CNN, YOLOv4, RetinaNet and . There is a maximum AP improvement of 1.9% and an average AP of 0.8% improvement on MS COCO dataset, compared to traditional evaluation-feedback modules. Here we just use as an example to illustrate the code.
Other
27 stars 6 forks source link

the Calculation of diou #4

Open life97 opened 3 years ago

life97 commented 3 years ago

Thank you very much for this interesting work. I added the diou to the IOU loss as a penalty item, but the network loss reached NaN during training？

Alan-D-Chen commented 3 years ago

The gradient disappeared. This means that the value of your loss function is too tiny. You need to coordinate the size ratio of each loss function. It's a result of trying. Please coordinate the value of each loss function to a unified order of magnitude!

Fateeeeee commented 3 years ago

Thank you very much for this interesting work. I added the diou to the IOU loss as a penalty item, but the network loss reached NaN during training？

I also met this stituation, do you solve it?