search

facebookarchive / C3D

C3D is a modified version of BVLC caffe to support 3D ConvNets.
Other
1.16k stars 506 forks source link

visuliaztion about C3D features? #123

Closed Thompshon closed 8 years ago

Thompshon commented 8 years ago

Hi, I meet a problem. We know you use the method from visuslizing and understanding convolutional networks to achieve the visualization of C3D models. Now, I don't exzactly understand how to visualize the 3D models using the method. I only can use the method achieve the visualization about picture. Could you tell me how to modify the source code realize the visualization of C3D models.

dutran commented 8 years ago

This implementation of C3D does not support visualization yet. We may support this int he future version.

Thompshon commented 8 years ago

image In the paper that Learning Spatiotemporal Features with 3D Convolutional Networks, we find the Figure4 that visualization of C3D model. Could you tell me how to get the similiar figure4.

zhuolinumd commented 7 years ago

@dutran can you provide some hints/ideas about how to visualize the learned C3D model? I thought you have implemented the deconv3d layer and unpooling3d layer. Can you share ideas/code? I think this is good for us to understand your approach. Thanks.