Modelpublic

kalpeshtiwari/efficientnetb0-steel-defect-detector

sourceHugging Faceupdated 1y agoView on Hugging Face

0likes21downloads

Model Card

EfficientNetB0 Steel Defect Detector

This is a binary classifier trained on Severstal Steel Surface images using EfficientNetB0.

—✅ Architecture: EfficientNetB0 with top 20 layers unfrozen
—🎯 Test Accuracy: 94.8%
—📈 AUC: 0.99

🔍 Use Case

Upload an image of a steel surface to see if it's classified as Defective or Non-Defective.

🖥 Live Demo

Try the Gradio demo on this page!

📦 Programmatic Access

python

from huggingface_hub import snapshot_download
import tensorflow as tf

model_path = snapshot_download("kalpeshtiwari/efficientnetb0-steel-defect-detector")
model = tf.keras.models.load_model(model_path)