Qustion about model runtime

[Noam-M]

I am curious to hear about the runtime and hardware that you used. You have stated in the paper that you used a desktop equipped with four NVIDIA 1080ti GPUs, but I did not see any runtimes stated. I am interested in inference time, as well as training times, in order to understand if we can use the architecture in real time applications.
Additionally, your paper presents a very interesting study of performance variations using different base networks, For example it states that the MobileNetV2- based model results in three times speedup when compared to the model based on the DenseNet-161. It would be very interesting to see the speedups for the other presented networks, as well as the inference time in absolute terms.

Much appreciated! Thanks!