Analyzing Failure Modes in Two-Stage HOI Detection Models
Explore key failure modes in two-stage human-object interaction detection models to improve accuracy in complex scenes and rare interactions.
KL-Based Quantization for Fast Mixed-Precision SSM-Transformers
Discover a KL divergence method for efficient, forward-only quantization in mixed-precision SSM-Transformer models, enabling faster edge AI deployment.
MaMe & MaRe: Efficient Token Merging for Vision Transformers
Discover MaMe & MaRe, matrix-based token merging and restoration methods that boost Vision Transformer speed with minimal accuracy loss.
Unified Conditional Flow for AI Motion Generation & Editing
Explore a unified conditional flow framework for AI-driven motion generation, editing, and intra-structural retargeting with improved accuracy and consiste...
Dynamic RGB-Event Fusion for Advanced Object Tracking
Discover MambaTrack's dynamic state transition and gated fusion for robust, real-time RGB-event object tracking with superior accuracy and efficiency.