Implementation of LTNet in "Facial expression recognition with inconsistent datasets", ECCV 2018
-
Data: Please download the cifar10 dataset from https://www.cs.toronto.edu/~kriz/cifar.html. For convience, we have uploaded our used images to OneDriver
-
Caffe: Please download the current CAFFE. Then, add the extra layers in folder "add to caffe" and recompile Caffe.
- Change the "source" and "root_folder" to the path in your computer.
- run "sh run_mixture.sh" to pretrain the model
- run "sh run_LTNet.sh" to train LTNet.
- Change the caffe_root to the path in your computer in EmotEngine.py and eval_cifar10.py
- Change Line 78,79 of eval_cifar10.py to the path in your computer
- run "python eval_cifar10.py -w path-to-model"