DeiT WITHOUT Distillation

This model classifies retinal images into:

  1. No DR
  2. Mild
  3. Moderate
  4. Severe
  5. Proliferative DR

Model architecture: deit_small_patch16_224

Distilled model: False

Primary metric: Tuned Quadratic Weighted Kappa

Primary inference output: Threshold-tuned QWK prediction

Tuned thresholds:

[ 1.0235428891379215, 1.7234984095439454, 2.025988978797374, 2.6660228022901524 ]

Downloads last month
17
Safetensors
Model size
21.7M params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Space using vyshnav112233/retina-dr-deit-without-distillation 1