HeiGIT released ~22M individual vehicle speed observations covering Europe's E-road network from 2022-2026, derived from Copernicus Sentinel-2 satellite imagery. Speeds are computed from a moving vehicle's echoes across the blue, green, and red bands, with per-vehicle trajectories, quality indicators, and timestamps shipped as GeoPackages for continental-scale traffic and transport studies.
New open dataset gives the Turkish AI community 309K+ high-quality Python instruction pairs, deduplicated and token-balanced to curb length bias in fine-tuning. A rare large-scale resource for instruction-tuning and evaluating Turkish-language coding models.
Italy's MIMIT ministry publishes every fuel station's prices daily as open data, each record timestamped with when the operator filed it, per a 2023 transparency decree. Analysts get a rare station-level, time-stamped series across 21,000+ stations, refreshed each morning and downloadable as CSV.
A free dataset of 183 non-financial Korean listed companies' balance sheets (FY2021 to FY2025), rebuilt from DART filings with a receipt number behind every figure, so any row traces back to its source document. Includes dual P/B (total vs controlling-interest equity), net cash, treasury shares and DPS. The CSV is email-gated; the table and JSON are open.