Day31 학습정리 - CV1

강의복습

1. Image classification 1

1) Course overview

2) Image classification

•  k Nearest Neighbors(k-NN)
• Convolutional Neural Netwoks(CNN)

3) CNN architectures for image classification 1

• AlexNet
• VGGNet
• GoogLeNet
• ResNet
• DenseNet, SENet, EfficientNet

2. Annotation data efficient learning

1) Data augmentation

• Brightness adjustment

• Rotate, flip

• Crop

• Affine transformation

• CutMix 

2) Leveraging pre-trained information

• Transfer Learning: 데이터셋이 적을 때 효과적
• Knowledge distillation: Passing what model learned to another smaller model(Teacher-student learning)

3) Leveraging unlabeled dataset for training

• Semi-supervised learning
• Self-training

---

코멘트

오늘 강의 내용은 3주차에 배웠던 내용을 조금 더 자세하게 다루는 느낌이었다.