반응형
Notice
Recent Posts
Recent Comments
Link
일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | ||||
4 | 5 | 6 | 7 | 8 | 9 | 10 |
11 | 12 | 13 | 14 | 15 | 16 | 17 |
18 | 19 | 20 | 21 | 22 | 23 | 24 |
25 | 26 | 27 | 28 | 29 | 30 | 31 |
Tags
- nfiti
- deep learning #segmentation #sementic #pytorch #UNETR #transformer #UNET #3D #3D medical image
- genetic epidemiology
- Surgical video analysis
- MICCAI
- non-parametric model
- words encoding
- nibabel
- parametric model
- TabNet
- parer review
- 데코레이터
- 유전역학
- decorater
- MRI
- TeCNO
- paper review
- Phase recognition
- monai
- parrec
- PYTHON
- precision #정밀도 #민감도 #sensitivity #특이도 #specifisity #F1 score #dice score #confusion matrix #recall #PR-AUC #ROC-AUC #PR curve #ROC curve #NPV #PPV
- tabular
- 파이썬
- 모수적 모델
- 비모수적 모델
- 코드오류
- 확산텐서영상
- nlp
- 확산강조영상
Archives
- Today
- Total
목록pytorch #torch #monai #dice #score #metric #segmentation #loss (1)
KimbgAI

Dice는 특이하게 loss function으로도 쓰이지만, metric으로도 사용된다. (그 이유는 segmentation이라는 task의 특수성 때문인데, pixel별로 class를 예측하기 때문에 metric score의 변화가 상당히 연속적이라 대부분의 구간에서 미분값이 유의미하기 때문이다.) (classification task에서 accuracy를 가지고 loss function으로 사용하지 않는 이유와 반대로 비슷하게 생각하면 된다.) 한편, 보통 Dice loss 는 1 - dice metric 으로 정의된다. 보시다시피 F1 score와 같다. 민감도(sensitivity)와 정밀도(precision)의 조화평균이다. monai에서 dice loss와 score를 편하게 계산할 수 있다...
machine learning
2022. 11. 8. 16:50