An Empirical Study of Automated Mislabel Detection in Real World Vision Datasets
Major advancements in computer vision can primarily be attributed to the use of labeled datasets. However, acquiring labels for datasets often results in errors which can harm model performance. Recent works have proposed methods to automatically identify mislabeled images, but developing strategies...
Zapisane w:
| Główni autorzy: | , , , , , |
|---|---|
| Format: | Artykuł |
| Wydane: |
2023
|
| Hasła przedmiotowe: | |
| Dostęp online: | https://arxiv.org/abs/2312.02200 http://arxiv.org/abs/2312.02200 |
| Etykiety: |
Nie ma etykietki, Dołącz pierwszą etykiete!
|