Justin Shenk comments

Results 46 comments of


                                            Justin Shenk

Padding only applied to first bounding box if needed

Thanks for catching this and opening the issue. Indeed, these should be fixed. I don't have time for a complete solution, but does https://github.com/justinshenk/fer/pull/34 look alright to you?

Padding only applied to first bounding box if needed

Thanks in any case for the quick response! On Wed, Oct 13, 2021 at 4:18 PM geloescht ***@***.***> wrote: > I guess that should work, unless always applying padding somehow...

Is there a flag to use CPU instead if GPU?

It's not currently possible in the library but try this: https://datascience.stackexchange.com/a/58846. On Thu, Jul 8, 2021 at 3:31 PM Fadwa ***@***.***> wrote: > — > You are receiving this because...

'Model' object has no attribute 'make_predict_function'

Which version FER are you using, what command did you run? On Sat, Mar 26, 2022 at 11:39 PM linguist89 ***@***.***> wrote: > I getting this error: 'Model' object has...

'Model' object has no attribute 'make_predict_function'

Can you show the outputs of `pip list`?

'Model' object has no attribute 'make_predict_function'

The FER version there is `21.0.6`. Maybe update the FER, the latest version is 22.4.0. On Wed, Mar 30, 2022 at 9:41 PM linguist89 ***@***.***> wrote: > I'm running it...

'Model' object has no attribute 'make_predict_function'

Can you share a minimal reproducible example? I can't reproduce this on my side. On Thu, Mar 31, 2022 at 12:01 PM linguist89 ***@***.***> wrote: > I've updated now to...

'Model' object has no attribute 'make_predict_function'

It seems to be rather the opposite, _make_predict_function is deprecated: https://www.tensorflow.org/api_docs/python/tf/keras/Model - thus your tensorflow version is somehow older. If you can share a minimal example I can look closer...

Model prediction is at 15 FPS, anyway to increase throughput?

Yes, we could batch images through the model. Are you using cpu only or also GPU acceleration?

Model prediction is at 15 FPS, anyway to increase throughput?

Consider setting the [`save_fps`](https://github.com/justinshenk/fer/blob/4ef7087050ddde00cf0fd8024ce08a8279ddd169/src/fer/classes.py#L174) parameter to a number 2-25. This will only compute on the nth frame.