Step 1:

Registration to the contest (since you have registered, you will receive all necessary data and will be informed of any updates, news etc.) (now open until 1/5/2007)

Step 2:

A training dataset toolkit (both for text line and word detection) will be sent to all registered participants (~ 20 document images + ground truth data + ground truth viewer & evaluation software) (just after registering)

Step 3:

All registered users is required to submit two executables (one for text line detection and one for word detection) in the form of Win32 console application (by 18/5/2007, extended)

Step 4:

The testing dataset (both for text line and word detection) will be sent to all registered participants (~ 100 document images + ground truth data) (on 25/5/2007, extended)

Step 5:

The participants with best working algorithms will be asked to submit a draft description of their text line and word detection methodology (by 1/6/2007, extended)

Step 6:

Final contest results to be announced at the ICDAR2007 conference