Face Image Annotation & Liveness Classification (Project)
ImageClassification
I developed a ViT-based biometric face liveness detector, which required collecting and annotating over 2,500 images for supervised learning. Data labeling involved identifying spoof versus real biometric facial images to train the anti-spoofing model. Model attention rollout was utilized to interpret and validate labeling accuracy.
• Labeled face image data for liveness classification.
• Used attention rollout to review and refine label quality.
• Supported model validation by iteratively relabeling as required.
• Ensured dataset integrity throughout the annotation process.
I developed a ViT-based biometric face liveness detector, which required collecting and annotating over 2,500 images for supervised learning. Data labeling involved identifying spoof versus real biometric facial images to train the anti-spoofing model. Model attention rollout was utilized to interpret and validate labeling accuracy. • Labeled face image data for liveness classification. • Used attention rollout to review and refine label quality. • Supported model validation by iteratively relabeling as required. • Ensured dataset integrity throughout the annotation process.
Not specified