Head Occlusion and Rotation State Image Dataset

Description: This dataset contains high-quality synthetic images depicting human heads under various occlusion conditions, including wearing hats, masks, sunglasses, and no occlusion, captured across multiple head rotation angles. It is generated using high-fidelity simulation environments to represent realistic facial and head movement scenarios. The dataset covers diverse demographic characteristics, lighting conditions, and background environments, ensuring applicability for general-purpose visual model training and head state perception system development.

Designed for face recognition preprocessing, head rotation modeling, and occlusion robustness testing, this dataset supports applications such as access control optimization, head pose estimation, and multi-angle identity verification. Each sample contains a high-resolution face image (1920×1080) with clear visibility of facial landmarks, head orientation, and occlusion category, enabling accurate training of multi-attribute recognition models.

The dataset comprises tens of thousands of images (exact count can be provided upon request) across multiple individuals and background scenes. Capture conditions simulate both indoor and outdoor natural lighting, with frontal and multi-angle head rotation coverage, including slight occlusion and full visibility scenarios. Each occlusion type—no occlusion, hat, mask, sunglasses—is represented across different head poses such as left turn, right turn, upward tilt, downward tilt, and frontal view.

All images are fully annotated with facial keypoints, head pose angles, and occlusion category labels, provided in JSON or COCO-compatible formats. Label distribution ensures balanced coverage across occlusion types and head poses, supporting robust algorithm development for pose-aware and occlusion-tolerant face recognition systems.

By combining photorealistic rendering, controlled occlusion variation, and detailed human annotation, this dataset provides an essential resource for developing accurate, resilient, and inclusive biometric and surveillance systems in real-world applications.

Keywords: head occlusion mask detection hat detection sunglasses recognition head pose estimation face detection

Sector: Biometrics / Surveillance / Access Control / Computer Vision

Metrics: Occlusion classification accuracy, head pose estimation error, face detection recall

Entities: Faces with hats, faces with masks, faces with sunglasses, unoccluded faces

Geographies: Global demographic representation

Price Range (USD): 4000–20000

Source type: Synthetic data generation from simulation environment, human annotation

Sample Images: