Abstract: Flow matching is a recent state-of-the-art framework for generative modeling based on ordinary differential equations (ODEs). While closely related to diffusion models, it provides a more ...
The forward and inverse kinematics networks of the data-driven path share a residual MLP backbone: MLP architecture for kinematics approximations. Inputs are projected into a 128-dim latent space and ...
The script below creates rolling-shutter images from a high-FPS source video and applies a point spread function (PSF) per source frame before row-wise sampling. python scripts/create_RS_image.py \ ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results