Manual extracting matches between frames

kukuruza / City-Project

Analyze traffic given a set of optical cameras in urban areas

0 stars 0 forks source link

Manual extracting matches between frames #29

Closed kukuruza closed 9 years ago

kukuruza commented 9 years ago

Motivation:

We'll need matches. Right now, we have only a few hundred.
Maybe, not all matches between a pair of frames is ok too
Matches from 100 frames takes 8 hours. However, it looks easy just to match already detected cars.

The idea is to create a fast interface to collect matches from frames with already detected cars. Not all cars are detected, so not all matches can be detected either. However, the result may bring so much new data.

satwikkottur commented 9 years ago

Do you propose to just show two detected cars (without the context) and ask for a match ?

kukuruza commented 9 years ago

yeah, just show two frames side by side, and let a user draw a line between two cars

kukuruza commented 9 years ago

Completed. Example to use is in learning/scripts/Manually.py Instructions at the description of function matchLabels() in learning/dbManual.py