toc: true title: Self Supervised Vision Transformers
tags: [‘temp’]
Self Supervised Vision Transformers
- An Empirical Study of Training Self-Supervised Vision Transformers
- recipes for Vision Transformer are yet to be built
- Self Supervised
- instability is a major issue that degrades accuracy, and it can be hidden by apparently good results
- improved when training is made more stable
- MoCO v3, a framework which offers an incremental improvement of MoCO
