[중간 점검] YOLOv5 객체 탐지 모델 v4 학습 결과 #6
yeverycode
started this conversation in
상황/결과 공유
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
현재 구축된 Roboflow Version 4 데이터셋을 기반으로 YOLOv5s 모델 학습 및 평가를 진행하였습니다.
데이터셋 현황
현재 포함 클래스
학습 환경
Validation/Test 결과
클래스별 mAP50
결과 해석
전체 mAP50은 0.885로, 10개 클래스 전체에 대해 비교적 안정적인 탐지 성능을 보였습니다.
braille, book_flower, tactile_stone 클래스는 높은 탐지 성능을 보였으며, book_stone과 book_flowerpot 클래스는 상대적으로 낮은 성능을 보여 추가 데이터 보강이 필요합니다.
산출물
향후 계획
data.yaml

metrics.txt
All reactions