Search papers, labs, and topics across Lattice.
cvlab, Yonsei University
6
0
6
Prompt-based continual learning can now capture diverse image distributions, overcoming the limitations of prompt collapse that hinder performance across tasks.
Middleware trade-offs in ROS 2 reveal that optimizing for spatial abstraction can compromise temporal guarantees, challenging the robustness of robotic communication.
Seam-sensitive adaptations in SHERPA enable the generation of 360掳 panoramas that maintain both photorealism and artistic style, overcoming traditional planar training limitations.
Traditional observation tools can inflate discovery costs and drop messages, but ros2probe achieves perfect recall while slashing resource usage by up to 28 times.
Robot middleware can transform into a critical "harness" layer that ensures safe and efficient integration of AI models into robotic systems by enforcing control, timing, and communication simultaneously.
DipGuava achieves photorealistic and personalized 3D head avatars from monocular video by explicitly disentangling facial appearance, outperforming prior methods in visual quality and quantitative performance.
