cvondrick / vatic

Efficiently Scaling Up Video Annotation with Crowdsourced Marketplaces. IJCV 2012
http://mit.edu/vondrick/vatic/
MIT License
607 stars 255 forks source link

Bug in annotation merging for attributes #8

Open cvondrick opened 12 years ago

cvondrick commented 12 years ago

From Nga Venkata Kartheek Medathati:

I am currently running vatic in offline mode. I annotated a video of approximately 3 minutes using action attributes after dividing the video into 3 segments (approximately 1500 frames each). When I dump the annotation into a text file using turkic dump command with merge option, I find that the action attributes of the last frames in the first segment are simply appended for the rest of the video segment but when I checked the attributes without merge option, the annotation is fine.