arXiv Computer Vision research summaries for March 20, 2024.

Today's Research Themes (AI-Generated):

• Text-to-3D shape generation advances but faces challenges in data representation and rendering.

• AutoBuild demonstrates efficient neural architecture search bypassing the costly traditional search process.

• Rotary Position Embedding enhances performance in Vision Transformers for various computer vision tasks.

• DetDiffusion synchronizes generative and perceptive models for improved data generation and perception in object detection.

• LaserHuman dataset and diffusion model set new benchmarks for language-guided, scene-aware human motion generation.