How to use Grad-CAM implemented by Gluon to visualize the model trained by symbol?

How to use Grad-CAM implemented by Gluon to visualize the model trained by the symbol network?

You could export your model and load it into Gluon. Here is an example:

net = gluon.nn.SymbolBlock.imports('resnet18-symbol.json',
                                   ['data'], 
                                   param_file='resnet18-0000.params',
                                   ctx=mx.gpu())

Once you have loaded the model into Gluon you can use the gradcam module. Here is a nice tutorial: https://mxnet.incubator.apache.org/versions/master/tutorials/vision/cnn_visualization.html

It is very helpful, thanks

The example is about HybridBlock. I didn’t find how to convert SymbolBlock to HybridBlock. Any advice will be appreciated, thanks

Sorry, I failed to try this. Any more advice will be appreciated, thanks

Hi @hdjsjyl,

did you get any solution for your problem? Would be interessted to see how you did it :slight_smile:

Thanks in advance!

Sorry, I didn’t get any solution. Thanks