FER+ stands for Face Expression Recognition Plus, which is an extension of the original FER dataset.

The dataset provides a new set of labels for the original dataset. Each image is annotated by 10 crowd workers, which not only provides better ground truth for facial expression recognition, but also enables researchers to more accurately estimate the probability distribution of facial emotions. These images are reclassified into eight emotion types: neutral, happy, surprised, sad, angry, disgusted, fearful, and contemptuous.