你好，论文中的公式5为了减少每个位置的重复计算，简化为使用所有神经元上的均值和方差，但是在实现代码中，方差除以了H*W-1。

I have read your paper, and I found there are regularizer is added, but not explain why adding this, and the introducing of regularizer will bring a hyper-parameter to adjust, this could make this method with a fixed hyper-parameter not generalize to other datasets.

δ²_t is not equal to 1/(M-1) * ∑(x_i - u_t)²,
it is equal to 1/(M-1) * ∑(x²_i - u²_t).

作者您贴出的类激活热力图有代码实现吗

The equation (3) should remove the minus sign

亲爱的作者您好，最近在查看您在百度云发布的mmdetection日志文件时产生了一些疑惑，望您解答。
下图是在您的百度云中下载的Mask RCNN Res50 的日志片段。如图所示，当验证集为coco2017val时，其总图片为5000，采用四卡验证应该显示数目为1250，但是您的训练日志验证时却显示的和训练时一样为14659。疑惑是否是实验设置时误把训练集也当作验证集了呢。

Hi,
you have done a splendid work to introduce the neuroscience finding to the machine learning area, and I have a question about the using of the energy function. In section 3.2, you proposed an energy function for each neuron to get the importances of the neurons, and by minimizing this function, you can get the linear transform which is the best transform to seperate the target neuron and other neurons.
However, I cannot understand your intention to use the minimum value of the energy function to represent the importance value. In my point of view, the energy function is kind of loss function, and the importance value should be the difference between

and

.
Therefore, why the energy function?
It would be very kind of you to answer my question.

Thanks,
Ema1997.

大佬您好，我把github中mmdection文件夹考到了我的mmdection文件夹中，并复制了提供的mask rcnn simam的log中model的config。但是运行的时候提示，ResNetRM is not in the models registry。请问想用mmdection运行您的代码还需要什么？求稍微给点关键词，如果有教程网站就更感激啦？
感谢！

Hi! Thanks for your code. I have a question about the Fast closed-form solution. How did you get this solution?

你好，我想问一下这个simam模块加在网络中的什么位置不会影响网络结构，并能取得不错的效果呢？还有个问题，在测试时候要用到吗？

作者您好！
感谢您的工作！
请问在simam中__repr__函数作用是什么？

    def __repr__(self):
        s = self.__class__.__name__ + '('
        s += ('lambda=%f)' % self.e_lambda)
        return s

谢谢！

谢谢作者的工作！菜鸟学生想提问几个问题，我用到了这两个文件加到了我的resnet骨干网络里面，我是做视频对象分割的，当我运行时出现了下列错误

不知道是不是当我将simam加入到bottleneck里面时，上下文数据处理维度的问题？

感谢感谢，烦请作者指点一二！

Hi, dear author!
Sorry to bother you.
Could you release the *.py of your visualized neuron(attention map/heatmap)?
Thank you!

你好，我在github中打不开预训练模型，您是否能够提供一下SimAM-R50预训练模型呢

It is a neural-free module?

Does the attention value self.activaton(y) needs to be self.activaton(y).detach()? Otherwise it will change the input value.

Did you try to convert onnx model or caffe model?

Welcome update to OpenMMLab 2.0

I am Vansin, the technical operator of OpenMMLab. In September of last year, we announced the release of OpenMMLab 2.0 at the World Artificial Intelligence Conference in Shanghai. We invite you to upgrade your algorithm library to OpenMMLab 2.0 using MMEngine, which can be used for both research and commercial purposes. If you have any questions, please feel free to join us on the OpenMMLab Discord at https://discord.gg/amFNsyUBvm or add me on WeChat (van-sin) and I will invite you to the OpenMMLab WeChat group.

Here are the OpenMMLab 2.0 repos branches:

Attention: please create a new virtual environment for OpenMMLab 2.0.

It seems like they are labels of target neuron t and other neurons x_i respectively. But what's the initial meanning of them in Eqn (1)?

Where is your requirements.txt?