search

dipika-singhania / ICC-Semi-Supervised-TAS

Iterative Contrast-Classify For Semi-supervised Temporal Action Segmentation
MIT License
9 stars 0 forks source link

dataset split #5

Closed hitcbw closed 2 years ago

hitcbw commented 2 years ago

Hi singhania, Thanks for your excellent work. I noticed that when generating semi-supervision on 50salads, the 2 or 3 samples in activity ID 1 first be chosen and another 23 samples in activity 2 are chosen. Why do you make this setting? Did I miss some explanation :)?

dipika-singhania commented 2 years ago

I didn't make this setting intentionally. It was random selection with videos from the 2 different types of salads mentioned in the dataset.