v1.2-experimental

Performance enhanced.

NOTE: some bug has been solved since this release, checkout master and use the trained models down here

Adaptive equalization (CLAHE) for better lighting handling.
Nice training with config 52x52, 1-5-8, svm.

Trained models 1vsall and 1vsallext. Download them and unpack (tar xvf svm_1vsall_1_5_8_95c2eb0b58.tar.gz; tar xvf svm_1vsallext_1_5_8_95c2eb0b58.tar.gz)

Build and run:

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..;make;make install
cd ../assets
./emotimegui_cli ../resources/haarcascade_frontalface_cbcl1.xml ../resources/haarcascade_eye.xml 52 52 1 5 8 svm ../svm_1vsall_1_5_8_95c2eb0b58/*
./emotimegui_cli ../resources/haarcascade_frontalface_cbcl1.xml ../resources/haarcascade_eye.xml 52 52 1 5 8 svm ../svm_1vsallext_1_5_8_95c2eb0b58/*

Performance summary of trained models:

Emotion	Accuracy
neutral	😄
sadness	😒
disgust	😄
anger	😄
surprise	😄
fear	😐
contempt	😒
happy	😄

experimental RC v1.1

Experimental version for demo purposes. Can be used with trained model at here and here

How to run the demo:

• download and unzip emotime-linux_x86_64-experimental-RC-v1.1.zip
• run ./download_trained_models.sh
• Then cd assets and
• ./emotimegui_cli ../resources/haarcascade_frontalface_cbcl1.xml none 48 48 3 5 4 svm ../dataset_svm_354_cbcl1_1vsallext/classifiers/svm/*