datdevsteve/nivravision-1.0-mvp

Nivra DINOv2 Vision (v2 refined)

Fine-tuned from facebook/dinov2-base on lesion vs normal/unclear images with strong normal dominance; Phase B unfreezes last 2 blocks.

Labels

{'infectious_skin_condition': 0, 'inflammatory_skin_condition': 1, 'normal_or_unclear': 2, 'suspicious_pigment_or_growth_lesion': 3}

Inference

from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import torch

model = AutoModelForImageClassification.from_pretrained('nivra_vision_dinov2_v2_refined')
processor = AutoImageProcessor.from_pretrained('nivra_vision_dinov2_v2_refined')
img = Image.open('your_image.jpg').convert('RGB')
inputs = processor(images=img, return_tensors='pt')
with torch.no_grad():
    logits = model(**inputs).logits
pred = logits.argmax(dim=-1).item()
print(model.config.id2label[pred])