MaskedFER2023 and MaskedCK+: Face-masked FER2013 and CK+ datasets for mask-aware facial expression recognition

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.

M-FER2013与M-CK+是FER2013和CK+数据集的掩码版本。CK+为小型数据集，因所含图像均取自视频的连续帧，故存在大量高度相似的样本。其来源为https://www.kaggle.com/datasets/shawon10/ckplus 与 https://github.com/spenceryee/CS229/tree/master/CK%2B。FER2013是另一款与CK+类似的面部表情识别数据集，规模更大，来源为https://www.kaggle.com/datasets/msambare/fer2013。M-FER2013_cropped与M-CK+_cropped分别为M-FER2013和M-CK+的裁剪版本：仅保留眼部周边区域，并剔除了掩码效果不佳（掩码尺寸与位置不合理）的样本。更多细节可参考https://github.com/hasan-rakibul/MaskTheFER。

搜集汇总

背景与挑战

背景概述

该数据集是MaskedFER2023和MaskedCK+，为FER2013和CK+数据集的带口罩版本，专为口罩感知的面部表情识别设计。它包括原始口罩版本及其裁剪版本，裁剪版本仅保留眼睛区域并移除不合适的口罩图像，以支持在口罩遮挡场景下的表情识别研究。