StaDy4D: Static Dynamic 4D Multi View Scene Dataset StaDy4D is the first large scale paired static–dynamic 4D dataset, comprising 9,000 paired sequences and ≈1.9M frames spanning diverse urban environments, camera configurations, camera types, and weather conditions. Each scene is captured with paired dynamic/static renders — identical camera trajectories with and without actors — generated from the CARLA simulator for 3D/4D reconstruction research. Car mounted Dynamic Static Drone Dynamic Static Orbit building Dynamic Static Orbit crossroad Dynamic Static CCTV Dynamic Static Pedestrian Dynamic Static TL;DR Paired static/dynamic renders from identical camera trajectories — enables scene decomposition and background reconstruction Per video : RGB (lossless MP4), depth map, camera extrinsics & intrinsics (safetensors) Ground truth : dense metric depth (0–1000 m), camera to world matrices, per frame 3D actor bounding boxes See Key Properties below for the full numerical breakdown. Quick Start Browse available scenes: Structure Splits Split City Maps Short/Mid Scenes Long Scenes Train City Map 1, 2, 4, 5, 6 270 85 Test City Map 3, 7, 8 162 51 Camera Types Camera Type Description cam 00…
Runs entirely in your browser via DuckDB-Wasm — this dataset's real data file is loaded once, then queried locally. Nothing is sent to a server.
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy