This paper introduces a hierarchical video summarization framework for long surveillance archives, leveraging scene clustering and salient motion cues to efficiently extract relevant segments and reduce monitoring workloads.

Andrew Simon McKenzie, Zhao Wenxin, Lavanya Meena Pillai, Christoph Andre Berger, Ying Yueqin, Daniela Elisabetta Romano

This study proposes a multi-modal fusion framework for disaster scene understanding, combining satellite, aerial, and ground-level visuals to provide accurate situational awareness and aid emergency response planning.

Franklin Robert Whitmore, Xu Lijun, Kavita Subramanian Das, Marco Giuseppe De Luca, Akira Masaru Saito, Eloise Charlotte Mayer

This paper introduces a gaze estimation framework using eye landmark regression and attention-aware modeling, enabling robots to infer human intent and enhance coordination in collaborative work environments.

Leon Gregory Matthews, Wu Yating, Sneha Anjali Raghavan, Pierre André Chavanel, Hyeonji Kang, Natalia Beatriz Alvarez

This study proposes a nighttime pedestrian detection system using fusion of thermal and visible images with context-guided attention modules, enhancing detection in low-light and cluttered urban scenarios.

Arthur James Caldwell, Zhang Qingyan, Ayesha Rashmi Menon, Sven Christian Bauer, Yui Riko Tanaka, Isabella Claire Moreno

This paper introduces a style-consistent feature embedding method combined with seasonal adaptation networks for robust visual localization across changing weather and lighting conditions in outdoor environments.

Isaiah Cole Whitman, Liu Zhihan, Meenakshi Shalini Batra, Giorgio Alessandro Conti, Marie Juliette Hossain, Chen Yuming

This paper proposes a vehicle re-identification framework that aggregates features across non-overlapping cameras using attention-based matching, enabling consistent tracking of vehicles in large-scale traffic surveillance systems.

Dominic Peter Walsh, Zhang Ruixuan, Anika Devika Sharma, Marcel Philippe Fournier, Yuna Haruka Matsuda, Gabrielle Sofia Ortega

This study introduces a prompt-based interactive segmentation framework using vision transformers, requiring only sparse user inputs to generate accurate object boundaries for augmented reality applications in real time.

Julian Edward Barrett, Huang Lixia, Bhavana Sudha Kiran, Pascal Antoine Chevalier, Hiroko Saki Nishimura, Camila Beatriz Rojas