Meteorite DINOv2-B Direct Model

Direct-call Hugging Face custom model for meteorite image classification.

from transformers import AutoModelForImageClassification
from PIL import Image

model = AutoModelForImageClassification.from_pretrained(
    "Eki734/meteorite-dinov2-b14-direct",
    trust_remote_code=True,
)
image = Image.open("example.jpg")
result = model.predict(image)
print(float(result["prob_meteorite"][0]))

This repository stores the trained MLP probe head. The frozen DINOv2-B with registers backbone (dinov2_vitb14_reg) is loaded from torch.hub at runtime.

Public score of the corresponding Top-90 submission: 0.79545.

Downloads last month
7
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support