This software lets you visualize Caffe models, specifically the convolutional kernels and weights for fully-connected layers of pre-trained models.

This program requires MATLAB and Caffe. It was tested on MATLAB R2015b and Caffe 1.0.0 RC3, but it should work for earlier versions of both.

1. Download the code (duh).

2. Move or copy startup.m.example to startup.m.

3. In startup.m, change the variable g_caffePath to the matlab folder under your Caffe installation.

4. (Optional) In startup.m, change the variable g_modelRootPath to the folder where you store your models. You don't necessarily have to set it to your model path, but since this determines the starting location of the GUI's file chooser (and nothing more), you should set this to a convenient location.

5. If needed, add the lib folder under your Caffe installation to your shared library path. For example, if the Caffe installation is at /home/rszeto/caffe, you would run the following command:

   export LD_LIBRARY_PATH=/home/rszeto/caffe/lib:$LD_LIBRARY_PATH
   

6. Open MATLAB. Make sure it has access to the shared library path above, which should be the case if you call matlab right after the above command.

7. Run startup.m.

Further instructions on using the GUI itself appear when you run the program.