search

coin-dataset / code

MIT License
47 stars 22 forks source link

Background in transcripts for weakly supervised action segmentation? #5

Closed atalnarayan closed 4 years ago

atalnarayan commented 4 years ago

Hi, For the weakly supervised segmentation task, do you include background in the ordered transcripts? For e.g., if a video has actions as (a_1, a_2, a_3) in respective order, but has background at the start, and in between a_2 and a_3, would the ordered transcript as per your paper's implementation be (b, a_1, a_2, b, a_3), or just (a_1, a_2, a_3)? Thanks in advance!

yzheng97 commented 4 years ago

The transcript should be (b, a_1, a_2, b, a_3). For the details of the annotation, you could refer to the Breakfast dataset: https://uni-bonn.sciebo.de/s/5ZEp7uWAVRQQMnh/download

yzheng97 commented 4 years ago

In breakfast, the background "step" is denoted as SIL.

atalnarayan commented 4 years ago

Hi, Thanks for your response! Yes, I am aware of the annotation in breakfast dataset, but wanted to confirm the inclusion of background in the transcript for experiments reported in the COIN paper since the json file does not have annotations for background regions.