Video Segments Annotation

This job is about labeling videos recorded by actors performing shoplifting actions.

Actors performed various actions during recording time. That includes shoplifting behaviors and regular behaviors.

Same behaviors were performed during same time. Behaviors are not mixed which means that same type of behaviors were recorded one by one. Timeline of behaviors were:
- putting things into pockets
- putting things into breast pockets
- putting things underwear (front and back)
- putting things into bags / backpacks
Then regular behaviors were recorded (these behaviors can be mixed):
examining items
- putting things into shopping basket
- examining prices
- putting items back

There are 10 videos 3h each. Each video contain about 200 actions of different types.

This job should be done VIA Video Annotator - experience with that software is required.

Selected action should be about 5-10 seconds long. Containing action performed with few seconds before and few seconds after (see example).

Suspicious/Shoplifting Behaviors:
in_pocket – Placing items discreetly into a pocket.
under_wear_front – Hiding items under the front section of a coat.
under_wear_back – Hiding items under the back section of a coat.
in_inner_breast_pocket – Placing items into an inner breast pocket of an unzipped coat.
in_sleeve – Slipping items into a coat or shirt sleeve. (this happens rarely)
in_bag – Hiding items inside a bag.
in_backpack – Hiding items inside a backpack.

Normal/Non-Suspicious Behaviors:
taking_item – Inspecting or handling items for evaluation.
return_item – Returning item on the shelf.
watching_on_shelf – Looking at item at shelf, not taking it

Example videos are available for people interested in that job.