layer

[mynameisvinn]

according to this: features = model.predict(frames_arr, batch_size=batch_size), you are extracting the output layer as feature representations. (in which case, thats incorrect: as youd want the last fully connected layer - that is, the layer preceding the output layer [1].)

[1] [Real-time Convolutional Neural Networks for
Emotion and Gender Classification

[Wazaki-Ou]

@mynameisvinn I'm not an expert (in fact, I just got the code today) but I think this line in the code here makes sure it does not extract the output layer:

    # Get outputs of model from layer just before softmax predictions

    from keras.models import Model
    model = Model(model.inputs, output=model.layers[-2].output)