This study presents a visual-inertial odometry solution tailored for indoor drones, utilizing lightweight sensor fusion networks and temporal consistency modeling to enhance flight stability and localization accuracy.

Jonathan Paul Whitaker, Zhang Ruihong, Aarthi Suman Reddy, Étienne Louis Charron, Haruka Miki Yamamoto, Clara Teresa Delgado

This paper introduces a generative approach for creating synthetic scenes in simulation environments, combining spatial priors and multimodal vision embeddings for realistic and diverse visual training data.

Leonard Francis Barrett, Liu Zhaoqiang, Shruti Manasa Iyer, Antoine Michel Perrault, Keiko Nanami Suzuki, Bianca Emilia Morales

This work presents a model for temporal segmentation of egocentric activities using multi-stage memory networks and semantic refinement modules, enabling accurate and efficient event parsing in first-person video streams.

Gavin Matthew O'Connor, Zhao Yuhan, Divya Pranathi Narayan, Laurent Nicolas Marchand, Emi Hoshiko Sakamoto, Lorena Carolina Pacheco

This paper proposes a contextual object tracking model for autonomous driving, incorporating multi-level attention and frame-adaptive learning to maintain robust performance in dynamic road scenarios.

Isaac Raymond Bennett, Zhang Qingli, Minal Sanjana Sharma, Jean-Marc André Pelletier, Yuna Riko Nakamura, Gabriela Alejandra Ortiz

This research introduces a visual question answering framework that applies scene graph reasoning with graph attention and language-guided traversal strategies to align visual understanding with natural language queries.

Bradley Simon Hughes, Li Wenxin, Rina Supriya Bhandarkar, Marcel Pierre Blanchet, Hana Aiko Fujimoto, Eliza Julieta Cruz

This study presents a cross-weather semantic segmentation model using style normalization and adaptive feature alignment, enabling autonomous vehicles to maintain perception accuracy in diverse weather conditions.

Julian Thomas McBride, Zhang Xiaoyan, Anushka Megha Ranganathan, Jean-Paul André Fontaine, Yuki Natsumi Sato, Mariela Cristina Varela

This paper introduces a spatiotemporal graph neural network for predicting pedestrian trajectories in traffic intersections, leveraging scene context and motion history for safe path planning in urban mobility systems.

Nathan Gregory Ellis, Zhao Minqing, Ishaani Raghavi Kapoor, Antoine Louis Renaud, Emi Sakura Tanaka, Isadora Lucia Mendes