ZEROTSUDIOS/cat-breed-classifier

🐱 Cat Breed Classifier

EfficientNetV2S fine-tuned to classify 67 cat breeds from a photo.

Model Details

Backbone	EfficientNetV2S (ImageNet pretrained)
Head	Dense(512) → Dense(256) → Softmax(67)
Input	224 × 224 RGB, scaled via preprocess_input
Classes	67 cat breeds
Training	2-phase: feature extraction then fine-tuning (top 80 layers)
Dataset	Cat Breeds — Kaggle

Usage

import tensorflow as tf
from tensorflow.keras.applications.efficientnet_v2 import preprocess_input
from huggingface_hub import hf_hub_download
from PIL import Image
import numpy as np
import json

# Download model
model_path = hf_hub_download("ZEROTSUDIOS/cat-breed-classifier", "cat_breed_model.h5")
model = tf.keras.models.load_model(model_path, compile=False)

# Predict
img = Image.open("your_cat.jpg").convert("RGB").resize((224, 224))
arr = preprocess_input(np.array(img, dtype=np.float32))
probs = model.predict(arr[np.newaxis], verbose=0)[0]
print(f"Top prediction: class {np.argmax(probs)} ({probs.max()*100:.1f}%)")

Live Demo

Built into a Streamlit web app — model is downloaded automatically on first run.