Skip to content
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

Support output for Global Pooling layers #28

Open
kretes opened this issue Nov 18, 2016 · 2 comments
Open

Support output for Global Pooling layers #28

kretes opened this issue Nov 18, 2016 · 2 comments

Comments

@kretes
Copy link
Contributor

kretes commented Nov 18, 2016

Global Average and Max Pooling are used as an alternative to FC layers.
they are very handy for vizualization. what quiver can do is show the convolution layers output that are input to GAP but GAP calculates average activation for a single channel of previous convolution.

it would be good to show this number alongside the visualized output of convolution layer if the next layer is GAP or GMP.

@kretes
Copy link
Contributor Author

kretes commented Nov 19, 2016

And furthermore there is yet another architecture where GAP is followed by a dense layer before given to softmax layer: http://cnnlocalization.csail.mit.edu/

supporting this as well would be really cool.

Actually those are the types of architectures that allow a built-in localization of classified item

@dan-1d
Copy link

dan-1d commented Nov 21, 2016

I agree that this would be a fabulous addition.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants