-
I tried to train the cf-net model using the code from the github and just replace the Mish activation function to Relu for the first 20 epoches and then back to Mish for another 15 epochs just as the…
-
my GPU is RTX 3090,so I have to use cuda 11, I already checked my cuda11.1 and it is useful,
but when I use : pip install git+https://github.com/thomasbrandon/mish-cuda/
to insall mish-cuda, there…
-
#CSPDarkNet
参考来源：
```
https://blog.csdn.net/bule_sky_wait_me/article/details/118547757
```
主要讨论YOLOv4中的backbone(特征提取网络)——CSP-DarkNet，以及其实现的所必需的Mish激活函数，CSP结构和DarkNet。
1. Mish激活函数
激活函数的目的是：提高网…
-
this is chart for pre-trained models using yolov4x-mish
This is chart for pr…
-
https://github.com/AlexeyAB/darknet/blob/05dee78fa3c41d92eb322d8d57fb065ddebc00b4/src/activations.c#L388
According to the formulation, the grad_sp should be caculated as follow;
const float grad_sp …
-
i want to activate yolov4 in spyder(ipython)
i install opencv python and opencv-contrib-python
next, i write code ( net=cv2.dnn.readnet(yolov4.weights, yolov4.cfg))
but, i have an error
non sup…
-
-
i trained yolov3 and yolov4 on my custom data using ultralic version. however, the mAP of yolov4(mish)is lower than yolov3 and yolov4(relu). But the paper mentioned yolov4 with mish activation functio…
-
-
Just a point of curiosity really. I noticed in your code that either ReLU or Swish is used. I understand the choice of including ReLU as it is commonly accepted as being very efficient. However the us…