Update (May 2019):
Check this new release: https://smoosavi.org/datasets/us_accidents. It contains more than 2.25 million traffic accident records which took place within the US, and will be periodically updated.
The following dataset is recently released and will be periodically updated: https://smoosavi.org/datasets/2019_traffic_weather_events
It covers ...
Open street map has incomplete data of building footprints for Qingdao. There are many ways to download OSM data, including:
Do a quick but messy download of all OSM data in a given area from the map interface. (That link will take you to the area of Qingdao, but you'll want to adjust the bounding box before starting the download.
Or check out the OSM wiki ...
You can try to get data from OpenStreetMap: the https://www.openstreetmap.org
Data can be downloaded per country or region on GeoFabrik.de: https://www.geofabrik.de/
There is even a nice python package called OSMNX to deal with road network (including bike ways where available): https://osmnx.readthedocs.io/en/stable/
You can select road with their OSM ...
Not per state, but here's the US consumption with various categories (on-highway, farm, etc)
opendata.dc.gov actually has some really nice datasets. DC is also a great choice for this, because it has so much good data over a very small area.
Code for DC has a lot of data too, but unfortunately their data portal seems to be down. You could join their Slack channel and ask them directly, or even stand up their portal yourself.
I'm not 100% they have ...
Many years later... (to anyone of interest), found this by chance, gyro, accelerometer vs long, lat for a study to predcit is the former one are more accuracte by using RNNs https://databank.illinois.edu/datasets/IDB-5975383