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
- cyclegan architecture
- lpips
- CPU
- 파이썬
- cyclegan loss
- 의료 용어
- cyclegan discriminator
- 자료구조
- 의학 기본 용어
- 역전파
- minmax loss
- cnn mlp
- 재귀함수
- CNN
- 단일 퍼셉트론
- GAN
- gan mode collapse
- 의료 영상 용어
- cyclegan generator
- 영상 의료 용어
- antslibrary
- cyclegan mode collapse
- mode collapse
- fsim
- cyclegan 구조
- 케라스 창시자에게 배우는 딥러닝
- cyclegan 구성
- mlp
- 모델성능평가지표
- alternative loss
Archives
- Today
- Total
JH 개발자의 성장일지
Deep Learning Study (21) / 24.12.24 본문
참고 : Registration — ANTsPy dev (latest) documentation
파라미터 리스트
- fixed - required
- moving - required
- type_of_transform - required
- initial_transform - optional
- outprefix - optional
- mask - optional
- moving_mask - optional
- mask_all_stages - optional
- grad_step - optional
- flow_sigma - optional
- total_sigma - optional
- aff_metric - optional
- aff_sampling - optional
- aff_random_sampling_rate - optional
- syn_metric - optional
- syn_sampling - optional
- reg_iterations - optional
- aff_iterations - optional
- aff_shrink_factors - optional
- aff_smoothing_sigmas - optional
- random_seed - optional
- write_composite_transform - optional
- verbose - optional
- multivariate_extras - optional
- restrict_transformation - optional
- smoothing_in_mm - optional
- singleprecision - optional
- kwargs - optional
Affine 다중 해상도 정합 과정
- Original Image: 고정(fixed) 및 이동(moving) 이미지를 입력으로 사용.
- aff_smoothing_sigmas: 각 해상도 단계에서 입력 이미지를 스무딩하여 노이즈를 제거하고 큰 구조에 초점을 맞춤.
- Downsample Image: aff_shrink_factors를 사용하여 축소해 저해상도 이미지를 생성.
- Compute Affine Transform: aff_sampling과 Mutual Information(MI)을 사용해 Affine 변환 매트릭스를 계산.
- Upsample Image: 고해상도로 이미지를 업샘플링.
- 고해상도로 복원하며 정합 작업 반복
Affine 정합과 관련된 파라미터
- 변환 방식, 정합 메트릭
- 샘플링 및 계산 설정
- 다중 해상도 설정
FOV 조정과 관련된 ANTs 파라미터
- type_of_transform : FOV 차이를 줄이기 위해 선형 변환(Affine) 또는 비선형 변환(SyN)을 선택
- aff_shrink_factors : 축소를 통해 고해상도 이미지를 저해상도로 변환하여 정합 속도와 효율성을 높임
- aff_smoothing_sigmas : 다중 해상도 단계에서 가우시안 스무딩 강도를 설정하여 노이즈를 줄이고, 큰 구조에 초점을 맞춤
- aff_sampling : Affine 정합에서 Mutual Information 계산 시 샘플링 비율을 조정 → FOV 차이가 크거나, 데이터 밀도가 높을 때 샘플링 비율을 높이면 정확도 향상
FOV 별도 전처리 메서드
- ants.resample_image
- ants.crop_image
- ants.pad_image
'Deep Learning' 카테고리의 다른 글
Deep Learning Study (22) / 24.12.29 (0) | 2024.12.31 |
---|---|
Deep Learning Study (20) / 24.12.21 (0) | 2024.12.31 |
Deep Learning Study (19) / 24.12.16 (0) | 2024.12.16 |
Deep Learning Study (18) / 24.12.12 (0) | 2024.12.12 |
Deep Learning Study (17) / 24.12.09 (0) | 2024.12.09 |