Gradiant places eighth in CVPR 2019 facial recognition anti-spoofing challenge

Gradiant has finished in the top 10 of the CVPR (Computer Vision and Pattern Recognition) 2019 international competition for biometric facial recognition technologies detecting spoof attacks with multimodal information.

Teams made up of more than 300 biometric technology experts from around the world participated in the challenge, according to the announcement, applying the latest advances in artificial intelligence to create robust and accurate anti-spoofing algorithms.

Gradiant was eighth overall, and second among entrants from Europe. The company will also present its research on “Deep Anomaly Detection for Generalized Face Anti-Spoofing” at the CVPR 2019 Congress, held June 16 to 20 in Long Beach, California. The proposed anti-spoofing system uses deep learning techniques to analyze RGB images for anomaly detection without processing the whole video.

The ChaLearn Face Anti-spoofing Attack Detection Challenge @CVPR2019 is based on a large-scale multi-modal dataset (Chelearn CASIA-SURF[1]) which is reported to be the largest publicly available dataset for anti-spoofing in terms of subjects and visual modalities. It has 21,000 videos of 1,000 subjects, with RGB, depth, and IR modalities for each sample.

“Presenting our solution in the CVPR, beyond the position obtained in the competition, is a recognition of the quality of Gradiant’s technologies,” says Gradiant Director of Multimodal Information Daniel González.

Gradiant also launched an automatic forgery detection technology for digital documents earlier this year.

Article Topics

accuracy  |  biometric testing  |  biometrics  |  Gradiant  |  spoof detection