PICO label correction algorithm implemented for Computer Vision
- Correct labeling is essential for image classification algorithms based on supervised machine learning.
- Nonetheless, labeling being a task that can only be done by people, human error is inevitable.
- PICOCV performs semi-automatic label correction for better performance.
- Install picocv via pip
pip install picocv
Example use case is demonstrated in "example" folder.
- DEVIEW 2019 발표자료 : 레이블링 조금 잘못돼도 괜찮아요: Clova가 레이블 노이즈 잡는 법