train_val 실험결과 공유(Cascade r-cnn) #75
kmouleejunhyuk
started this conversation in
Show and tell
Replies: 2 comments
-
오오 차이가 굉장히 적네요 |
Beta Was this translation helpful? Give feedback.
0 replies
-
해당 데이터셋으로 retinanet을 돌려본 결과입니다. 제출 성능: 0.471 cascade r-cnn보다 차이가 조금 나네요.. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
결론: valid와 submission 차이가 크지 않은 좋은 valid set인 것 같습니다.
실험결과: https://wandb.ai/cider6/detr/runs/159tbiob?workspace=user-torque
데이터셋 경로: /opt/ml/detection/mmconfig/sample/cascade_rcnn_trainval_test/
브랜치: train_val_test
EDA상 큰 문제가 없어 학습 수행
제출 성능: 0.453
bbox_map_75: 0.37 -> delta 0.083
bbox_map_50: 0.456 -> delta 0.003
감사합니다
Beta Was this translation helpful? Give feedback.
All reactions