Text Locating

Entry Instructions: Details here

The aim of this competition is to find the system best able to locate text regions in scenes.  Downloads now available.

Example:

We expect the text locating algorithm to find the following text regions:

That is, five rectangles in total.  Ignore the small text above each rectangle, this is the ground truth text for the robust reader, and not used for the text locating competition.

Text Locater Interface:  the text locater is expected to take as input an image, and produce as output a set of rectangles, where each rectangle bounds a region of text.

Note that the text locater is not expected to read the text - merely find regions that contain text.

Downloads

Sample and Trial Dataset now available