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 |
- pdf merge
- Google API
- YOLOv7
- yolo
- google cloud
- YOLOv5
- DeepLearning
- objectdetection
- 파이썬 #업무자동화 #python
- computervision
- Text To Speech
- pyautogui
- Text-to-Speech
- processstart
- ironpython
- 업무자동화
- 사무자동화 #Selenium
- 사무자동화
- pypdf2
- pythonnet
- Today
- Total
Doarchive
GAN 모델의 종류 DCGAN , LSGAN , Conditional GAN 본문
DCGAN (Deep Convolutional Generative Adversarial Network)
딥러닝 모델인 CNN을 사용하여 생성자와 판별자를 구성
생성자는 무작위 노이즈 벡터를 입력으로 받아 진짜같은 이미지를 생성하고, 판별자는 생성된 이미지와 실제 이미지를 입력으로 받아 각각을 구분하도록 훈련됨
생성자는 판별자의 결과를 최대화하고, 판별자는 생성자와 실제 이미지를 구분하도록 훈련되며, 이 과정에서 생성자는 실제 이미지와 비슷한 이미지를 생성하게 됨
DCGAN은 고해상도 이미지를 생성하며, 이미지 내의 객체나 패턴 등의 특징을 잘 포착함.
이미지 간의 변환, 예를 들면 블랙앤화이트 이미지에서 컬러 이미지로 변환하는 작업에서도 사용됨
논문
https://arxiv.org/abs/1511.06434
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between
arxiv.org
TensorFlow 공식 DCGAN 예제:
https://www.tensorflow.org/tutorials/generative/dcgan
심층 합성곱 생성적 적대 신경망 | TensorFlow Core
심층 합성곱 생성적 적대 신경망 컬렉션을 사용해 정리하기 내 환경설정을 기준으로 콘텐츠를 저장하고 분류하세요. Note: 이 문서는 텐서플로 커뮤니티에서 번역했습니다. 커뮤니티 번역 활동
www.tensorflow.org
PyTorch 공식 DCGAN 예제
https://github.com/pytorch/examples/tree/main/dcgan
GitHub - pytorch/examples: A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc.
A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc. - GitHub - pytorch/examples: A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc.
github.com
LSGAN (Least Squares Generative Adversarial Network)
판별자의 출력값과 실제 값의 차이를 최소화하는 L2 손실 함수를 사용하여 생성자와 판별자를 훈련시키는 방법
기존 GAN에서 발생할 수 있는 모드 붕괴(mode collapse)와 같은 문제를 완화시킬 수 있다
논문
https://arxiv.org/abs/1611.04076
Least Squares Generative Adversarial Networks
Unsupervised learning with generative adversarial networks (GANs) has proven hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. However, we found that this loss function may lead to t
arxiv.org
Conditional GAN
조건부 생성적 적대 신경망(Conditional GAN)
생성자와 판별자 모두가 조건을 입력으로 받아 이미지를 생성하거나 분류를 수행하는 방법
예로 입력 이미지가 고양이인지 개인지에 따라 생성된 이미지가 달라지도록 할 수 있다.
이 방법은 이미지 생성 뿐만 아니라 이미지 변환, 이미지 분할 등의 분야에서도 활용
논문
https://arxiv.org/abs/1411.1784
Conditional Generative Adversarial Nets
Generative Adversarial Nets [8] were recently introduced as a novel way to train generative models. In this work we introduce the conditional version of generative adversarial nets, which can be constructed by simply feeding the data, y, we wish to conditi
arxiv.org
'Computer Vision > Deep learning' 카테고리의 다른 글
Faster R-CNN (0) | 2023.04.18 |
---|---|
DC-GAN(Deep Convolutional Generative Adversarial Network) (0) | 2023.04.14 |
YOLOv7 사용 방법 (0) | 2023.04.05 |
YOLOv7 Error : subprocess.CalledProcessError: Command 'git tag' returned non-zero exit status 128. (0) | 2023.04.05 |
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x99 in position 7884: invalid start byte 해결 방법 (0) | 2023.03.28 |