Are there benchmarks for assessing the speed of the forward-pass of neural networks?

Question

I have a task where I would like to use a convolutional neural network (CNN). I would like to incrementally start from the fastest models, fine-tune and see whether they fit my "budget". At the moment, I'm just looking at object detection CNN-based feedforward models.

I'm curious to know if there is any article, blog, web page or gist that benchmarks the popular CNN models based on the forward-pass speed. If there is back-propagation time and dataset-wise performance, even better!

score 1 · Accepted Answer · edited Nov 04 '19 at 03:34

1

https://github.com/jcjohnson/cnn-benchmarks might be a good start. It mostly focuses on GPUs, but there is also one CPU (Dual Xeon E5-2630 v3).

edited Nov 04 '19 at 03:34

nbro

39,006
12
98
176

answered Jul 05 '18 at 12:24

cantordust

943
6
10

Are there benchmarks for assessing the speed of the forward-pass of neural networks?

1 Answers1