-
Notifications
You must be signed in to change notification settings - Fork 173
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Training Steps Mismatch in the paper and the code in ImageNet Experiments #24
Comments
this is my PyTorch implementation of CSPDarknet. I borrow some functions from mmdetection and mmcv.
|
My PyTorch code is posted on #24 (comment). I am sorry about that I can not release my lightweight models due to some issues. |
@WongKinYiu in AlexeyAB/darknet#3708 (comment) while in AlexeyAB/darknet#5355 (comment)
what's more,in AlexeyAB/darknet#5355 (comment) I want to know which |
@nyj-ocean Hello,
|
@WongKinYiu |
the latest and in my previous experiments, i used sam layer as: |
@WongKinYiu |
yes, it is same. |
@WongKinYiu |
Hi, I have checked the network structure and number of parameters in my CSPResNet/CSPResNeXt PyTorch implementation, which is the same as what you reported in your Github README file, including nn.BachNorm2d, nn.LeakyReLu, Training epochs, batch size and learning rate schedule. I also have a close look at your DarkNet PyTorch implementation. However, the ACC point is still below yours... My Results:
Thanks! |
I am not sure it is important or not, I just follow https://pjreddie.com/darknet/imagenet/. And I think gets a little bit lower accuracy is normal, since darknet use 256x256 for validation, and I guess your PyTorch code use 224x224 instead. Could you share your code of CSPResNet / CSPResNeXt, I would like to upload the implementation and results to pytorch branch if it is OK. |
@WongKinYiu I notice that the modified SAM in yolov4 paper is reference to the CBAM paper. However, I also find that ThunderNet paper also design a SAM. so I want to know:
|
There are many kind of channel attention module (CAM) spatial attention module (SAM) in the literature. For example SENet and SKNet proposed different kind of CAM, and CBAM and ThunderNet prposed different kind of SAM. In general, we will cite the first paper or the most similar paper or both in related work. So the answer of your question is:
No, they are different.
The CBAM is the first paper which proposed SAM, we cite it in yolov4 paper. The ThunderNet prposed the most similar SAM module as ours, we cite it in cspnet paper. |
@WongKinYiu |
yes, all of different kind of sam modules produce the attention of spatial. |
@WongKinYiu |
Hi @WongKinYiu Thanks for your reply! I think that during training and testing, the DarkNet framework keeps the image size as 256256. However, for common PyTorch training, the training size is 224224, and the test size is 256*256. Is my understanding right? |
it is depend on your code. |
@WongKinYiu |
i do not know too, i always use the anchors which yolo9000 calculated. |
@WongKinYiu @AlexeyAB |
|
@AlexeyAB |
@AlexeyAB I find that there are many black spare parts in my own
|
i guess images in your dataset are form videos. |
What is the black spare? |
@AlexeyAB there are many black spare parts in my own
|
@WongKinYiu |
Because your objects are small relative to the image size. This is normal. Just may be you should use higher network resolution for anchors calculation, training and detection to get good results. https://github.com/AlexeyAB/darknet#how-to-improve-object-detection
|
@AlexeyAB |
@AlexeyAB |
@Chaimmoon Could you share me your code of CSPResNet50? |
I'm sorry to bother you again. I have another question about yolov4 paper modify SAM from
|
|
Thanks for your reply. what I understand about
These questions are very troubling to me. I look forward to your answers.Thanks a lot |
Hi,
In ImageNet Experiments, the paper said that it should be trained for 800 epochs:
However, in the code, it said that it should be trained for 80 epochs:
So there is a big difference……
Besides, I try to re-implement in PyTorch, and the ACC is 7~8 points behind your method. The network architecture and number of parameters is the same as your Darknet results……
Best,
Mu
The text was updated successfully, but these errors were encountered: