A research team including Amazon has released GOFish, a method that builds a full-field-of-view 3DGS and a well-aligned mesh from fisheye images alone. It aims to produce simulation-ready digital twins without LiDAR.
Why it mattersRobots, autonomous vehicles, and 360° cameras often use fisheye lenses. For workflows that build reconstructed real-world environments from captured footage, being able to use the full field of view out to the edges could be an advantage. However, whether it works in real-world settings depends on what code and models are released and on testing with your own data.
Ziang Guo and colleagues propose SG^2SLAM, a 3DGS SLAM for outdoor monocular navigation. They say it outperformed conventional frame-by-frame correspondence methods on KITTI and the Waymo Open Dataset.
Why it mattersIf a dense map and trajectory can be obtained from a monocular camera alone, the method could be used for outdoor navigation on small mobile platforms and robots that have difficulty carrying LiDAR. However, validation on real hardware is not confirmed in the source material.
A method that controls 3DGS density using a score derived from the reconstruction loss, rather than heuristics, has been released. Code for standard 3DGS is on GitHub, and results on the seven Mip-NeRF360 scenes report higher metrics with fewer Gaussians.
Why it mattersIf quality metrics can be raised while reducing the number of Gaussians, this could help cut file size and rendering load. For practitioners working on web and mobile delivery and viewers, it could be one way to limit counts on the training side.