Requesting Faster rcnn model related details in action detection

[malarsaravanan1991]

Hi, We are working on action detection and our work closely resembles to your work “Rethinking Spatiotemporal Feature Learning:Speed-Accuracy Trade-offs in Video Classification”. We would like to cite your work in our paper. I would like to know the following model related details of your action detection on faster RCNN framework :

1. Number of trainable parameters in your model
2. GFLOP's(floating point operation) with both flow and RGB
3. Details of the GPU and inference fps. Any help is much appreciated