Wangmeng Zuo

New method improves learning new classes with less data.

ConSept framework enhances semantic segmentation by reducing forgetting in models.

A new method enhances depth maps using color details effectively.

A new tool for making images safer and more responsible.

IMWA enhances model performance in class-imbalanced learning tasks efficiently.

A new approach enhances the effectiveness of attacks on safety-focused language models.

CPT improves black-box model performance without direct access to internal parameters.

A new method for assessing text-to-video models focuses on dynamics.

A new method improves the process of creating smooth animation transitions.

AutoDirector simplifies film production through efficient task management and user involvement.

A new method transforms text into detailed 3D scenes seamlessly.

LPT++ improves object recognition in classes with few examples through advanced techniques.

Ensuring robots can perform tasks without causing harm or chaos.

A new system ensures consistent multi-view videos for better self-driving car training.

Discover new methods in machine learning for better class recognition.

RHFL+ tackles data noise and model differences in federated learning.

A new method for creating smooth video transitions with Frame-wise Conditions-driven Video Generation.

Models learn old and new objects while remembering past knowledge.

MMD-LoRA helps autonomous vehicles estimate depth during challenging weather conditions.

MetricDepth enhances depth estimation from single images using deep metric learning.