High loss, high accuracy

[Dirac99]

Dear Sir,

I have been trying to reproduce the results and got some issues: 1.) High loss in age classification, >0.54 on average, the accuracy is comparable to your results. 2.) Very high accuracy in gender classification, 0.96 on average. I am not sure what is wrong. I have attached the log file for gender_fold_0 below.

gender_fold_0.txt

Thank you in advanced.

[GilLevi]

Thank you for your interest in our work.

The log file is the output of the training process? Or are you trying to test the accuracy after training? I'm asking since it didn't look to me like a training output, but I might be wrong. Can you please explain?

Best, Gil

[Dirac99]

I am testing the accuracy using caffe model from "cnn age gender models and data.0.0.2" on the website. And I run the code from terminal:

./build/tools/caffe test -model AdienceFaces/AgeGenderDeepLearning-master/gender_net_definitions/train_val_test_fold_is_0.prototxt -weights AdienceFaces/cnn_age_gender_models_and_data.0.0.2/gender_net.caffemodel -gpu 0 -iterations 100

Ps: I am totally new in caffe and machine learning, sorry if I miss out information. Thank you.

[GilLevi]

Hi,

I never used Caffe's command line test tool. I recommend using the python wrapper to test the code. You can find an example of using our model with python here: http://nbviewer.jupyter.org/url/www.openu.ac.il/home/hassner/projects/cnn_agegender/cnn_age_gender_demo.ipynb

Best, Gil.