New gait retrieval task and dataset introduced by researchers

Researchers have introduced Composed Gait Retrieval (CoGR), a new task focused on retrieving gait sequences based on a reference sequence and a natural language modification query. To support this task, they developed two new gait-language datasets, Language-Augmented CCPG and Language-Augmented CASIA-B, using large vision-language models for annotation. They also proposed ComposeGait, a framework designed to maintain identity during retrieval by using a Part-aware Identity Adapter to aggregate identity evidence into ID tokens, which are then injected into a shared Q-Former to prevent identity drift. ComposeGait achieved state-of-the-art results on both newly created benchmarks. AI

IMPACT Introduces new methods for gait recognition using language, potentially improving human identification systems.

RANK_REASON The cluster describes a new research paper introducing a novel task, dataset, and framework. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• CASIA-B dataset
• Composed Gait Retrieval
• ComposeGait
• FAM120B
• Language-Augmented CASIA-B
• Language-Augmented CCPG
• Part-aware Identity Adapter
• Q-Former

• paper
• other

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