food-recognition-model

Food Recognition Model (UECFOOD256)

This model is a deep learning classifier capable of recognizing 256 different types of food. It was trained using TensorFlow/Keras on the UECFOOD256 dataset.

Model Details

• Model Architecture: [INSERT BASE MODEL HERE, e.g., MobileNetV2 / EfficientNetB0] (Fine-tuned)
• Dataset: UECFOOD256 (256 food categories)
• Framework: TensorFlow / Keras
• Task: Image Classification (Multi-class)

Intended Use

This model is designed to take an image of food as input and output the specific food category name. It is useful for:

• Calorie tracking apps
• Dietary monitoring
• Food logging automation

How to Use

You can load this model directly in Python using the huggingface_hub library.

Important: This model requires a class_names.txt file (included in the repo) to map the model's integer predictions back to readable food names.

Inference Code

import numpy as np
from huggingface_hub import from_pretrained_keras, hf_hub_download
from keras.preprocessing import image
import tensorflow as tf

# 1. Define the Repo ID
repo_id = "Hiratax/food-recognition-model"

# 2. Load the Model
print("Loading model...")
model = from_pretrained_keras(repo_id)

# 3. Load the Class Names Mapping
class_names_path = hf_hub_download(repo_id=repo_id, filename="class_names.txt")
with open(class_names_path, 'r') as f:
    class_names = [line.strip() for line in f.readlines()]

# 4. Prediction Function
def predict_food(img_path):
    # Resize image to 224x224 (Standard for this model)
    img = image.load_img(img_path, target_size=(224, 224))
    img_array = image.img_to_array(img)
    img_batch = np.expand_dims(img_array, axis=0)

    # Predict
    predictions = model.predict(img_batch)
    predicted_index = np.argmax(predictions[0])
    confidence = np.max(predictions[0])
    
    return class_names[predicted_index], confidence

# 5. Run on an image
# food, conf = predict_food("my_sushi.jpg")
# print(f"Prediction: {food} ({conf:.2f})")