SplashSplat reconstructs splashing liquids from multi-view videos
SplashSplat: Reconstructing Splashing Liquids from Real-World Multi-View Videos
Computer Vision and Pattern RecognitionGraphics
Summary
Reconstructing splashing liquids is hard because they change shape very fast and look different from every angle. The authors created a new set of videos showing real splashes from seven cameras to help with this problem. They also made SplashSplat, a method that builds 3D liquid shapes and motions only where the video data supports it. This method produces smoother liquid movement and better results than previous techniques. It can even create smooth transitions between moments and apply new visual styles without starting over.
What this means in practice
- •For visual effects artists: Create realistic 3D liquid splashes from multi-camera recordings for use in film and animation.$Commercial implications: Enables production studios to generate accurate splash effects from video, reducing manual animation effort.
- •For robotics perception engineers: Improve understanding of fast-moving liquids in multi-camera setups for better robotic interaction and manipulation.
Authors
Peiyu Liu, Dingxi Zhang, Federico Tombari, Marc Pollefeys, Christina Tsalicoglou, Daniel Barath
Abstract
A splash lives for a fraction of a second: sheets tear into ligaments and droplets, appearance is view-dependent and nearly textureless, and little persists long enough to track. Reconstruction research has consequently focused on smoke, synthetic liquids, or gently deforming surfaces. To our knowledge, no synchronized multi-view dataset of splashing liquids exists. We therefore introduce a benchmark of 20 real scenes, from coherent streams to violent splashes, captured by seven synchronized, calibrated 4K cameras at 60 fps, with manually refined per-view liquid and container masks and fixed evaluation splits. We further present SplashSplat, built on a single principle: impose physical structure only where the observations can constrain it. Per-frame liquid SDFs fused from the masks provide the geometry, level-set transport between consecutive SDFs yields a coarse velocity field, and Lagrangian carriers advected along this flow, corrected against each new observation and reseeded where coverage is lost, decode local Gaussians for differentiable rendering. SplashSplat outperforms state-of-the-art dynamic Gaussian splatting methods on our real captures and on a synthetic benchmark, with physically more plausible motion and a lower training cost. The same representation supports temporal interpolation and style transfer without re-optimization.