search

oarriaga / face_classification

Real-time face detection and emotion/gender classification using fer2013/imdb datasets with a keras CNN model and openCV.
MIT License
5.61k stars 1.59k forks source link

Issue with training dataset #43

Closed depakaakula closed 7 years ago

depakaakula commented 7 years ago

Do I have to let it train 10000 times really?

screen shot 2017-11-19 at 13 56 22
oarriaga commented 7 years ago

No, the model checkpoint will stop saving the models whenever the validation loss function stops decreasing.

depakaakula commented 7 years ago

Thanks. Is it possible to generate the output file for whatever is getting logged in terminal?

Thank you.

On Mon, Nov 20, 2017 at 04:42, Octavio Arriaga notifications@github.com wrote:

No, the model checkpoint will stop saving the models whenever the validation loss function stops decreasing.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/oarriaga/face_classification/issues/43#issuecomment-345642466, or mute the thread https://github.com/notifications/unsubscribe-auth/Ae_KFHIXdSjbsL7vrWAVTLBkYBbl8AYgks5s4UmhgaJpZM4QjECd .

oarriaga commented 7 years ago

There is a callback from keras that allows you to save the losses in a csv file. I believe the current training script contains such callback and is generating the log files while running.

depakaakula commented 7 years ago

Dear oarriaga. The log file (fer2013_emotion_training.log) doesn't seem to be capturing my recent terminal report. Also, the output file data is kinda unstructured. Can you help me understand how to relate the expression and digits in the output file?

Thank you.

screen shot 2017-11-30 at 00 21 22