search

shamangary / SSR-Net

[IJCAI18] SSR-Net: A Compact Soft Stagewise Regression Network for Age Estimation
Apache License 2.0
584 stars 146 forks source link

Inference time #21

Closed zhly0 closed 5 years ago

zhly0 commented 5 years ago

Hi，shamangary： how did you evaluate your cpu time？I use the following code to evaluate the inference time： start_CPU = timeit.default_timer() results = model.predict(faces) end_CPU = timeit.default_timer() in the TYY_demo_mtcnn.py code，but it takes about 20-30ms in cpu time，is it correct？

shamangary commented 5 years ago

Hello @zhly0. I don't have the exact number now since the time depends on your computer hardware, but it seems correct. Note that the slow part of the demo is the face detection part. So you need a fast face detection to achieve real-time.

zhly0 commented 5 years ago

@shamangary . I test attribute model only，it seems to ten times slower than your model(in your paper,the cpu inference time is 2.69ms),I do not know if the hardware makes the difference.