EfficientDet.Pytorch
EfficientDet.Pytorch copied to clipboard
Performance of model efficient D0
Follow this paper, I saw efficient D0- Flops is smaller than yolov3 28 times. I think the speed of this model is higher than yolov3 ~ 28 times. Is this right? I calculated the performance on GTX-2080 : Yolov3-Alexab/darknet ~ 0.027 - 0.029 s/frame -> ~30frame/s Your git - efficient-net D0 ~ 0.027s/frame -> ~30frame/s This speed is same, so have any problem in here? Can you explain it to me? By the way, thank for your hard work 👍
@buiduchanh I think there are 2 causes that effect performance:
-
- According to paper, Efficientdet-D0 use backbone B0, Wbifpn = 64, Dbifpn = 2. In my repo, t changed it to Wbifpn = 256, Dbifpn = 1 => It will affect performance much. Anyway, I will try public weights + performance like paper soon.
-
- If you yolov3 on darknet with inference c++. We cann't compare performance with this method because my repo code by python
why did you change the hyperparameter on EDet-D0? is there any particular reason?
I'd appreciate it a lot if you could tell me how to calculate the fps?