Advancing Facial Age Progression for Occluded Faces

It is observed that face recognition is highly vulnerable when the age gap between the gallery and the probe images is drastically high. This phenomenon is a universal concern since acquiring gallery images at multiple age intervals might not always be possible. Therefore, accurate age progression is an ideal solution to mitigate this age gap and boost face recognition performance. It is observed that the existing age progression algorithms are vulnerable to occlusion. Keeping this in mind, this paper presents a novel approach to facial age progression, particularly addressing the challenge of occluded faces. The objects occluding the face's key points are first detected using segment anything and later inpainted using transformer architecture to improve the age progression. We compare our results against state-of-the-art models across various age clusters (e.g., 0-3, 15-19, and 50-69), demonstrating superior performance in terms of age progression and retaining identity, gender, and age attributes. The proposed work significantly improves facial age progression's robustness and visual quality, enhancing its applicability in security systems, forensic analysis, and other fields requiring precise age prediction.