Styl3R predicts stylized 3D Gaussians in less than a second using a feed-forward network given unposed sparse-view images and an arbitrary style image.

MBA-SLAM can accurately estimate the local camera motion trajectories within the exposure time and recovers the high quality 3D scene from blurry sequences.

We explore the possibility of recovering the neural radiance fields and camera motion trajectory from a single blurry image with its events.

BAD-Gaussians successfully deblurs severe motion-blurred images, synthesizes higher-quality novel views, and achieves real-time rendering, surpassing previous SOTA implicit deblurring rendering methods.

We address the challenge of detecting object defects through the self-supervised learning approach of solving the jigsaw puzzle problem.

USB-NeRF can remove rolling shutter effect and recover high-fidelity high frame-rate global shutter video from a sequence of rolling shutter images.

BAD-NeRF jointly learns the 3D representation and optimizes the camera motion trajectories within exposure time from blurry images and inaccurate initial poses.
Nov. 2023 update: we have released the NeRFstudio version Bad-RFs.

BAD-NeRFstudio can shorten the training of BAD-NeRF to minutes.