deep-emotion-recognition-src

train_image.py : This is the file for creating a model using transfer learning using one of previously trained networks.
train_image_oneVres.py : This file is for generating expert models for each emotion.
train_image_merge_models.py : This file is for stacking models using a two layer MLP.
train_image_ptest_sample.py : This file is like train_image.py but for being run multiple times from run_train_image_ptest_sample.sh
train_image_merge_ptest_sample.py : This file is like train_image_merge_models.py but for being run multiple times from run_train_image_merge_ptest_sample.sh

Extract the dataset archive. In train_image.py do following steps:

Point DATA_DIR to location of dataset.
Specify WEIGHTS_FILE to one of inception, resnet, vgg16 or vggface. Path to a hdf5 file which contains network weights with network structure can be specified, as well.
Specify a WORK_DIR location of wish.
And run.
Results will be written under WORK_DIR. Models will be saven under WORK_DIR/checkpoints directory. WORK_DIR/logs directory will cotain training history.
Confustion matrices, validation results, accuracy by epoch chart will be generated.

In order to prepare an Ubuntu system for deep learning follow instructions:

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
bash_scripts		bash_scripts
docs		docs
extract		extract
face		face
pickle_files		pickle_files
thirdp		thirdp
util		util
work_image		work_image
work_video		work_video
README.md		README.md
__init__.py		__init__.py
exctract_features_fer13_numpy.py		exctract_features_fer13_numpy.py
main_extract.py		main_extract.py
train_image.py		train_image.py
train_image_merge_models.py		train_image_merge_models.py
train_image_merge_ptest_sample.py		train_image_merge_ptest_sample.py
train_image_oneVres.py		train_image_oneVres.py
train_image_ptest_sample.py		train_image_ptest_sample.py

habanoz/deep-emotion-recognition-src