Controllable Person Image Synthesis

Controllable person image synthesis aims to generate realistic human images with precise control over pose, shape, appearance, and even clothing. Recent research focuses on developing 3D-aware generative models, often employing neural radiance fields or vertex-based representations coupled with techniques like attention mechanisms and Gaussian splatting to achieve high-fidelity results and handle complex clothing deformations. These advancements are driving progress in virtual try-on, avatar creation, and animation, impacting fields like computer graphics, virtual and augmented reality, and digital fashion. The ability to generate highly realistic and controllable human images has significant implications for various applications.