-2

I am working on classification problem for Covid19, Pneumonia and healthy lung.

I have 3000 images for each class.

  • Can I apply vision transformers for this image classification instead of normal CNN?
  • Or is there any prerequisite for applying this? I am new to transformers.

I have tried all CNN and they have achieved 95% accuracy till now.

Minsky
  • 2,277
  • 10
  • 19
vamsi bharadwaj
  • 27
  • 5

1 Answers1

2

At small to medium datasets, ViTs don't give a performance that is comparable. On really big datasets, however, they have outperformed CNNs. More information here https://www.v7labs.com/blog/vision-transformer-guide

Enos jeba
  • 72
  • 1
  • 3
  • 10