AxonData/pc-web-ir-rgb-webcam-dataset-ibeta-level-1

iBeta Level 1 PAD — IR + RGB Webcam Face Liveness Dataset for PC/Web Application

Summary

iBeta Level 1 paper attacks captured with a dual-camera setup: IR (infrared) + RGB webcam for PC/web-based onboarding. Each sample contains a synchronized IR channel and a standard webcam RGB view, covering bona-fide recordings and four paper spoof variations: print & cutout, cylinder, on-actor, and 3D paper masks. Designed for ISO/IEC 30107-3 PAD Level 1 pre-evaluation in a PC/Web onboarding flow — web application via desktop webcam (IR + RGB)

Full version of dataset is availible for commercial usage - leave a request on our website Axonlabs to purchase the dataset 💰

This dataset targets paper-based presentation attacks that are typically evaluated in iBeta Level 1 testing. All samples are recorded simultaneously in two camera modalities:

• RGB webcam (PC/web) — a standard desktop/laptop webcam that mirrors real web-onboarding scenarios
• IR camera (infrared) — an IR channel captured in parallel to enable IR-based liveness and multi-modal fusion (IR + RGB) The focus is on PC / web-based attacks: a subject sits in front of a monitor and interacts with the camera like in a browser-based KYC or workforce login flow. Every attack type is available both in IR and in RGB

Camera & Recording Setup

• Dual-camera capture: IR and RGB webcam streams for each sample
• Environment: desktop/office setting to emulate PC/web-based identity verification
• Active liveness motion: sequences include natural movements (e.g., slight head turns, zoom-in / zoom-out) to reflect real active checks
• Content types: videos multiple paper mask spoof variations

Attack Taxonomy

• Print & Cutout — printed face with cutouts for eyes/mouth (print/cut)
• Cylinder — curved / cylindrical print to simulate facial volume
• On Actor — a flat paper mask worn by a live performer (on-actor) with eye/head variations
• 3D Paper Mask — volumetric paper masks with protrusions (e.g., nose) or mannequin mounting All attack types are captured from the RGB webcam (PC/web) and from the IR camera, enabling cross-modal training

Potential Use Cases:

• Liveness detection (PAD): train and evaluate algorithms that separate bona-fide webcam selfies from paper-based spoof attacks under PC/web conditions using RGB, IR, or IR + RGB fusion
• Pre-iBeta evaluation: stage models against iBeta Level 1 like paper attacks before formal certification