Dataset are scraped from Instagram Public Data. Lots of thanks to creators in Hong Kong on Instagram.

Weighting from 0.6 ~ 0.8. This can be used with other LoRa like KoreanDollLike.

Please use it reasonably.