07: Emerging Properties in Self-Supervised Vision Transformers (DINO)

Self Supervised Learning
Representation Learning
一种无需标签的自监督学习方法，通过教师–学生自蒸馏训练 Vision Transformer，自发涌现出语义一致的全局表示与清晰的注意力分割能力。

1 Preliminary

2 DINO

3 Summary

4 Key Concepts

5 Q & A

Back to top