On Matching Faces with Alterations due to Plastic Surgery and Disguise
This addresses a critical problem in biometric security and surveillance by improving face recognition accuracy under extreme alterations, though it is incremental as it builds on existing deep learning methods.
The paper tackles the challenge of face recognition under plastic surgery and disguise variations by proposing a framework that transfers fundamental visual features from a generic image dataset to supplement a supervised model, achieving state-of-the-art results, including over 87% verification accuracy at 1% false accept rate on the DFW dataset, which is 53.8% better than baseline.
Plastic surgery and disguise variations are two of the most challenging co-variates of face recognition. The state-of-art deep learning models are not sufficiently successful due to the availability of limited training samples. In this paper, a novel framework is proposed which transfers fundamental visual features learnt from a generic image dataset to supplement a supervised face recognition model. The proposed algorithm combines off-the-shelf supervised classifier and a generic, task independent network which encodes information related to basic visual cues such as color, shape, and texture. Experiments are performed on IIITD plastic surgery face dataset and Disguised Faces in the Wild (DFW) dataset. Results showcase that the proposed algorithm achieves state of the art results on both the datasets. Specifically on the DFW database, the proposed algorithm yields over 87% verification accuracy at 1% false accept rate which is 53.8% better than baseline results computed using VGGFace.