search

YijinHuang / SSiT

SSiT: Saliency-guided Self-supervised Image Transformer for Diabetic Retinopathy Grading
23 stars 4 forks source link

compare the result with a plain backbone. #13

Closed osivaz61 closed 8 months ago

osivaz61 commented 8 months ago

I am a newbie. I am trying to understand the process. Did you compare the results of the plain backbone (supervised model like in Lesion-Based Contrastive Learning) with the proposed model? Are the results of the plain model or the proposed model high? There is no comparison.

YijinHuang commented 8 months ago

Thank you for your question. Random init. is the supervised plain model, which is a plain model that randomly initialized trained on downstream datasets.