3D Gaussian and Diffusion-Based Gaze Redirection

FlowLensing is a new model introduced for simulating gravitational lensing, which is crucial for studying dark matter. This model utilizes a Diffusion Transformer-based approach to efficiently create high-fidelity lensed images. FlowLensing significantly enhances simulation speed, achieving over 200 times faster performance compared to traditional methods, while maintaining physical consistency across various dark matter models and parameters.