Masked Thermal Infrared Face Image ((MaskTIF)

MaskTIF is collected from different scenarios from ETIF and contains 3,000 thermal infrared face images. The collection time was during the severe COVID-19 epidemic period, and contains a large number of mask-wearing samples. There are no constraints impose on pedestrians during the data collection process, the dataset may include various facial images, which range from faces with glasses or masks to side faces, as illustrated in Fig. 5. Since our SelfFS is a self supervised method, the collected infrared face images can be directly used to train the model without human annotated. In order to test the performance of the model, we constructed a fever infrared face test set with reference to the method in DGDC [14]. Furthermore, half of the infrared faces in each period are employed for training, and the remaining ones are utilized for testing.