happy8825/siglip-ecva-main

SigLIP Classification Model - main

Pushed: 2025-12-16T15:38:25.488860Z

Metrics

• Best (accuracy): 0.6731601731601732 @ step None
• Final eval: accuracy=0.579004329004329, f1=0.5481997677119629

Train Sampling

• mode: balanced
• before: normal=56738 abnormal=30844 total=87582
• after: normal=30844 abnormal=30844 total=61688

Inference

from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import torch
processor = AutoImageProcessor.from_pretrained('happy8825/siglip-ecva-main')
model = AutoModelForImageClassification.from_pretrained('happy8825/siglip-ecva-main')
img = Image.open('your_image.png').convert('RGB')
inputs = processor(images=img, return_tensors='pt')
with torch.no_grad():
    logits = model(**inputs).logits
pred = logits.argmax(-1).item()
print(model.config.id2label[pred])