The code of HSEmotion Python Library is released under the Apache-2.0 License. There is no limitation for both academic and commercial usage.

    python setup.py install

It is also possible to install it via pip:

    pip install hsemotion

    from hsemotion.facial_emotions import HSEmotionRecognizer
    model_name='enet_b0_8_best_afew'
    fer=HSEmotionRecognizer(model_name=model_name,device='cpu') # device is cpu or gpu
    emotion,scores=fer.predict_emotions(face_img,logits=True)

The following values of model_name parameter are supported:

• enet_b0_8_best_vgaf
• enet_b0_8_best_afew
• enet_b0_8_va_mtl
• enet_b2_8
• enet_b2_7

The method predict_emotions returns both the string value of predicted emotions (Anger, Contempt, Disgust, Fear, Happiness, Neutral, Sadness, or Surprise) and scores at the output of the last layer. If the logits parameter is set to True (by default), the logits are returned, otherwise, the posterior probabilities are estimated from the logits using softmax.

In addition, it is possible to extract visual embeddings for classifier learning

    features=fer.extract_features(face_img)

The versions of these methods for a batch of images are also available

    emotions,scores=fer.predict_multi_emotions(face_img_list,logits=False)
    features=fer.extract_multi_features(face_img_list)

Complete usage example is available in the test_hsemotion_package.ipynb. It is necessary to run the following line to run the demo script:

run pip install facenet-pytorch

The details about training of the models are available in the main repository