MaskedFER2023 and MaskedCK+: Face-masked FER2013 and CK+ datasets for mask-aware facial expression recognition
Description
M-FER2013 and M-CK+ are the masked versions of FER2013 and CK+ datasets. CK+ is a small dataset that contains many highly similar images, as these images are consecutive frames from videos. Source: https://www.kaggle.com/datasets/shawon10/ckplus and https://github.com/spenceryee/CS229/tree/master/CK%2B. FER2013 is another facial expression recognition dataset similar to CK+ with a larger size. Source: https://www.kaggle.com/datasets/msambare/fer2013 M-FER2013_cropped and M-CK+_cropped are the cropped versions of 'M-FER2013' and 'M-CK+', respectively. Only regions near the eyes are preserved, and some poorly masked images (inappropriate masked sizes and positions) are removed. Refer to https://github.com/hasan-rakibul/MaskTheFER for more details.
Files
Steps to reproduce
They are obtained by running the MaskTheFace algorithm on FER2013 and CK+ datasets. The implementation of MaskTheFace can be found at https://github.com/aqeelanwar/MaskTheFace. You will need to install MaskTheFace and follow the instructions in its readme file. Additionally, it is important to overwrite its utils/aux_functions.py file with the one provided under the /src/ folder at https://github.com/hasan-rakibul/MaskTheFER, where we have made some modifications to the parameters to ensure that the algorithm can work on low-resolution grayscale images.