Computer Vision Section 058
Training Vision Models
How to take a model somebody else trained and make it work on your few thousand photos.
12 of 12 lessons published Three reading levels on every lesson
Lessons in order
Work top to bottom. Each lesson assumes the one above it.
- Fine-tuning a vision model
- Freezing and unfreezing layers
- Replacing the classifier head
- Linear probing vs fine-tuning
- Where ImageNet pretraining stops helping
- Few-shot image classification
- Domain adaptation for vision
- Geometric augmentations
- Mixup and CutMix
- RandAugment and AutoAugment
- Building an Albumentations pipeline
- Test-time augmentation