Thermal Imaging Face Detection Based on Transfer Learning

Thermal facial recognition is getting more and more attention now. However, due to the small number of thermal imaging facial databases, models trained with existing databases are not as robust and have low recognition rates in different environments. To solve this problem, we released a new thermal surface database and added the SE attention mechanism module and SPP module to the YOLO v3 framework to improve the detection effect of the model. Compared with the original YOLO v3, the mAP value has been improved by 22.1 % and 16.7%.