Nong Sang

TetherMem enables video generation models to dynamically adapt scenes while keeping subjects consistent, achieving a significant leap in overall video quality and scene progression.

CEAVAD redefines anomaly detection by transforming the inference process into a competition between hazard and benign event hypotheses, leading to superior performance without training data.

CSNet achieves state-of-the-art image fusion by intelligently contrasting clarity and selecting optimal pixels, redefining how we approach multi-focus image challenges.

CMAM revolutionizes text-based person retrieval by achieving granularity-aware matching that significantly outperforms state-of-the-art methods across diverse query levels.

By harnessing clean latent features, SharpMoE dramatically improves resource allocation to salient tokens, overcoming a key limitation in diffusion MoE architectures.

A novel graph-based dataset pruning method achieves over 40% faster training on ImageNet-1k while maintaining accuracy, challenging the status quo of sample selection.