AgeNet age_list ranges

idata · ‎03-09-2019

Hello there!

I have a quick question, refering to the age list of AgeNet run.py file:

age_list=['0-2','4-6','8-12','15-20','25-32','38-43','48-53','60-100']

Why there is no age 3, 7, 13, 14, 21 to 24 etc…? How is these left out 'ages' to be accounted using this network? Is there any way can get the ages accurately by changing different model or mean file?

Thank you :smile:

idata · ‎03-12-2019

Hi @RickyChew

The ages in between those ranges probably weren't included in the training of the model. If a person's actual age falls in between those ranges, they will be put into the next category. For example, for a person that is 22 years old, it will predict their age in between 25-32 years old with x% certainty.

OpenVINO has pre trained age & gender model here that is able to give more exact results (instead of ranges). However, the training of this model did not include children, so it's able to recognize people in age ranging from 18-75.

Please let me know if this was helpful.

Best Regards,

Sahira

idata · ‎03-13-2019

Hi @Sahira_at_Intel

Thank you very much for reaching out!

That's great to hear. I have been using NCS 1 since mid 2018 before OpenVINO and NCS 2 was introduced to me earlier this year and most of the environment I am currently using is based on NCSDK and NCS 1. However, in my future project I would love to migrate into NCS 2 with OpenVINO with improved examples, library, documentations etc.