3

I want to write an application which takes a single image an classifies each pixel as "street" or "not street". To do so, I need training / testing data. One possible application is for self-driving cars or car assistance systems. So the camera is within the car and looks at the street from a drivers perspective.

Where can I get such (preferably labeled) data?

(I have found KITTI, but the site seems to be down at the moment and more data is always better.)

1 Answer 1

2

You may do it a hard way. YouTube and other video hostings have a bunch of DVR videos that capture lots of real-world events.

Find videos on YouTube

Say, by the Russians:

http://www.youtube.com/results?search_query=russian+dvr

Take random screenshots from these videos

https://stackoverflow.com/questions/26850636/youtube-api-grab-screenshot-of-a-video-at-a-specific-time

Label it with Amazon MTurk

https://www.mturk.com/mturk/welcome

Then do the analysis.

You can also ask data from authors who published on your issue and may have well-behaving datasets.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.