What is transfer learning in Computer Vision?
Updated May 15, 2026
Short answer
Transfer learning reuses pretrained models for new vision tasks.
Deep explanation
Pretrained models like ResNet or EfficientNet trained on ImageNet are fine-tuned on smaller datasets. Early layers capture generic features like edges, while later layers are retrained for task-specific patterns.
Real-world example
Using ImageNet pretrained models for cancer detection.
Common mistakes
- Training entire model from scratch on small datasets.
Follow-up questions
- What is fine-tuning?
- Why freeze layers?