Computer Vision Architecture Evolution: ConvNets to Transformers - Lecture 21

Computer Vision Architecture Evolution: ConvNets to Transformers - Lecture 21

UCF CRCV via YouTube Direct link

Improvements

10 of 24

10 of 24

Improvements

Class Central Classrooms beta

YouTube videos curated by Class Central.

Classroom Contents

Computer Vision Architecture Evolution: ConvNets to Transformers - Lecture 21

Automatically move to the next video in the Classroom when playback concludes

  1. 1 Introduction
  2. 2 Evolution of Vision Architectures
  3. 3 Hierarchy of SWIN vs. CNNs
  4. 4 Modernizing ConvNets
  5. 5 Modernizing ResNet
  6. 6 Macro Design Changes
  7. 7 Changing stage compute ratio
  8. 8 Changing stem to "Patch-ify"
  9. 9 Depthwise Conv. vs Self-Attention
  10. 10 Improvements
  11. 11 Inverted Bottleneck
  12. 12 Larger Kernel Sizes
  13. 13 Micro Designs (mD)
  14. 14 Replace RELU with GELU
  15. 15 Fewer Activation functions
  16. 16 Fewer Normalization Layers
  17. 17 Substituting BN with LN
  18. 18 Visualization
  19. 19 mD4- Improvement
  20. 20 Separate Downsampling Layer
  21. 21 Final ConvNext block
  22. 22 Networks for Evaluation
  23. 23 Training Settings
  24. 24 Machine Performance Comparison

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.