Questions about the face-recogintion sample which demonstrates Plugin API for TensorRT2.1


 I found the face-recognition sample using TensorRT2.1 provided by you on github:"". I want to compare the original caffe code of the plugin layer with the code written in pluginImplement.cpp to learn how to write the plugin layer. However the webpage doesn’t mention the name of the face-recognition network.And where can I find the caffe version of this network? Thanks.

Hello 393702504,

I don’t know if this helps, but by looking at the “deploy.prototxt”, I can tell the network they are using looks like 2 GoogleNet ( merged together.


We merge a DetectNet and a GoogleNet to be the face-recognition network.

The training Caffe branch is NVCaffe.
The implemented plugin layer focus on the concatenation between two networks rather than a particular layer inside it.
As a result, we don’t need any custom layer to train each model but only apply the concatenation when running them together.

We share lots of information on this topic. You can check it for more information.