大纲

  1. Pruning-based
    • DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsificatio
    • IA-RED$^2$ : Interpretability-Aware Redundancy Reduction for Vision Transformers
    • EVIT: EXPEDITING VISION TRANSFORMERS VIA TOKEN REORGANIZATIONS
    • ATS: Adaptive Token Sampling For Efficient Vision Transformers
  2. Different computational cost
    • Dynamic Grained Encoder for Vision Transformers
    • Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer
  3. Early exitting
    • Not All Images are Worth 16x16 Words: Dynamic Transformers for Efficient Image Recognition
    • Coarse-to-Fine Vision Transformer

      视频链接

      腾讯会议

      PPT链接

      百度网盘 提取码:dhbo