New OAPR method uses rare attributes for person retrieval in surveillance

Researchers have introduced Open-Attribute Person Retrieval (OAPR), a new method for finding individuals in surveillance footage based on specific, often rare, attributes rather than just common descriptors. This approach is designed to handle attributes like 'holding a weapon' or 'lying on the ground' that can significantly narrow down search results. To support OAPR, a new dataset called EPAD was created, featuring over 267,000 images and 65 distinct attributes. The proposed GAP-CLIP framework, a lightweight CLIP-based system, demonstrates strong performance in retrieving individuals based on these open-ended attribute queries, including those not seen during training. AI

IMPACT This research could improve the efficiency and effectiveness of person identification in surveillance systems by enabling more precise attribute-based searches.

RANK_REASON Academic paper introducing a new method and dataset. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• GAP-CLIP
• Minjeong Park
• Open-Attribute Person Retrieval

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →