Mapping Crosswalk Coverage via Satellite Imagery in San Francisco.
I particularly commend the absolute grind of the data collection method:
The method was time-consuming, but simple. I manually reviewed satellite imagery at every single intersection citywide. I did this in order to maximize accuracy, and dig deeper than a binary crosswalk vs. no-crosswalk score.
did he (try getting the info from|upload them to) OSM?
Why don't you ask someone who would know, like for example, not me?
Or you could read the article -- or even the pull-quote here -- for why he decided that trusting someone else's pre-digested map data was not a good way to get accurate results.
Nice to see that there authors recognize that an unmarked crosswalk is still a crosswalk. If most motorists also expected people to walk across unpainted intersections then the lack of marking would not be such a big problem.