by Jeong-Yoon Lee
Inspired by Max Woolf’s benchmark, the performance of 3 different backends (Theano, TensorFlow, and CNTK) of Keras with 4 different GPUs (K80, M60, Titan X, and 1080 Ti) across various neural network tasks are compared.
For the performance of TensorFlow and CNTK with K80, the numbers reported at Max Woolf’s benchmark are used.
The accuracies of Theano, TensorFlow and CNTK backends are similar across all benchmark tests, while speeds vary a lot.
Among K80, M60, Titan X and 1080 Ti GPUs:
Detailed results are available at https://github.com/szilard/benchm-dl/blob/master/keras_backend.md