L
Laya Vision
A 256M model that answers typed questions about an image with calibrated odds
Emerged finds new products worth a look, daily — what’s new today
DRS-VPT
Finds where a single camera image was taken inside a LiDAR or indoor 3D map
Robotics and mapping teams can place a camera in an existing scan from one image, with no starting guess or known intrinsics, using one model indoors and out.
DRS-VPT is the code and weights behind an ECCV 2026 paper by Lanke Frank Tarimo Fu and Maurice Fallon. Given one image and a point cloud, it estimates the camera's full 6-DoF pose. The same model handles driving LiDAR scans (KITTI, nuScenes), indoor maps and iPhone photos, and it can calibrate a camera to a LiDAR without targets. Code is Apache-2.0, but the 2.9 GB of weights are non-commercial only. It runs on Linux with an NVIDIA GPU, or on CPU.
More like this
B
Black Hole Raytracer
Fly a camera past a Schwarzschild black hole and watch light bend, in a browser
BlindSight
Webcam person detection where the cloud server never sees a single frame
CoordMap
Geocoding, UTM and MGRS, area and sun times in one page, no sign-in
Burst Super-Res
Stack several shots of a still scene into one cleaner photo at double size
Low Poly Earth
The whole planet as a stylised 3D model you can fly over and photograph on a Mac