Design an annotation format for videos

[filipbartek]

Design a data format for annotating recordings of slinky juggling. The format should be able to describe the position of the slinky (each of its ends) and the hands in every frame. The slinky ends and the hands should be differentiated.

[filipbartek]

Partially implemented in FrameAnnotation

[filipbartek]

This would be useful namely for

• using machine learning to improve object detection (annotated recordings as samples) and
• testing (annotated recordings as test cases).

[filipbartek]

Preferable format: a text file that accompanies a video file.

[filipbartek]

Partially implemented in 7d0fdc3.
The program now stores all detected positions with time stamps in a CSV file. These could (although probably not easily) be coupled with video frames and as such visualized or analyzed.
As of now, the output CSV file can be used for example for prototyping of beat estimation method in R.