Model Description

SkinSAM is on the 12-layer ViT-b model, the mask decoder module of SAM is fine-tuned on a combined dataset of ISIC and PH2 skin lesion images and masks. SkinSAM was trained on an Nvidia Tesla A100 40GB GPU.

Some of the notable results taken:
ISIC Dataset:

  1. IOU 78.25%
  2. Pixel Accuracy 92.18%
  3. F1 Score 87.47%

PH2 Dataset:

  1. IOU 86.68%
  2. Pixel Accuracy 93.33%
  3. F1 Score 93.95%
Downloads last month
48
Safetensors
Model size
93.7M params
Tensor type
F32
ยท
Inference API
Inference API (serverless) does not yet support transformers models for this pipeline type.

Dataset used to train ahishamm/skinsam

Space using ahishamm/skinsam 1