Hanna Paulava comments

Results 6 comments of


                                            Hanna Paulava

is it possiple to use this code to extract the feature with my own trained model

@Saira05 - you probably forgot to add /models/research/slim to you PYTHONPATH, see README setup 2nd step https://github.com/tomrunia/TF_FeatureExtraction#setup

My streaming json string throws an error in my jasonlite fromJSON R script function

Could you reproduce this error on your local machine?

Overfitting

@NikitaKononov I can share my training graphs. I'm currently at 0.29 train/eval loss. Not overfitted yet, but nearly there. But still better than 0.39. What I did - really strictly...

Overfitting

> * Yep! My loss also similar! Also I took 6M to get below at 0.25 Thanks! In this case there's still 2.5 million steps to go :sweat_smile: the loss...

Add best practices for working with dates

Hi @antoine-182! Yes, if you have some ideas, I still didn't finish it. Contribution is warmly welcome!

Code not working

Actually `demo_syncnet.py` has a pretty restricted usage. It will work only for videos that are similar in format (i.e. size, amount of faces) to `example.avi`. In order to obtain results...