Cant get the accuracy value (classification) from facial recognition program

Hi,

I am using a facial recognition example in nvidia (https://devtalk.nvidia.com/default/topic/1023699/jetson-tx2/questions-about-face-recongnition/post/5209485/#5209485). However, I can’t get the accuracy value for classification from the facial recognition. I have read some of the post and it said is the res for the classification and for detection box is conf.

I tried printing out the res value but it does not shows the % accurately. It either show 0% or 0.00000000012%.

Thanks!

Hi,

Do you use a custom model?
It’s recommended to check the accuracy of detection and classification model first.

After that, please also check the Recognition plugin if any modification required for your use case:
https://github.com/AastaNV/Face-Recognition/blob/master/pluginImplement.cpp#L293

Thanks.

Hi,

Yes I used a custom model, and I manage to get the correct reading for classification in RecognitionLayer::Summary.

Thanks!

Hi,

I have another question is that when I execute it, after 3 to 5mins it will auto close/clash with the error of Segmentation fault (core dumped)? Is it suppose to happen?

Thanks!

Hi,

You may find some relevant information here:
https://devtalk.nvidia.com/default/topic/1007313/jetson-tx2/how-to-build-the-objection-detection-framework-ssd-with-tensorrt-on-tx2-/5

Thanks.

Hi!

Have read it and the problem is solve! thanks so much!

Good to know this. : )