App/dog_vision.py

import tensorflow as tf
import tensorflow_hub as hub
import numpy as np
import os
import pandas as pd
import cv2
import keras
from matplotlib.pyplot import imread
from flask import jsonify

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

# Create a function to load a trained model
def load_model(model_path):
    '''
    Loads a saved model from a specified path.
    '''
    print(f'Loading saved model from: {model_path}')
    model = tf.keras.models.load_model(model_path,
                                        custom_objects={'KerasLayer': hub.KerasLayer})
    return model

loaded_full_model = load_model(r'C:\Users\piyus\OneDrive\Desktop\FlaskApp\Model\final-keras-model-xception.keras')



labels_csv = pd.read_csv(r'C:\Users\piyus\OneDrive\Desktop\FlaskApp\Model\labels.csv')
labels = labels_csv['breed'].to_numpy()

# Find the unique label values
unique_breeds = np.unique(labels)
IMG_SIZE = 150

# Turn probabilities into their respective label (easier to understand)
def get_pred_label(prediction_probabilities):
    '''
    Turn an array of prediction probabilities into a label.
    '''
    return unique_breeds[np.argmax(prediction_probabilities)]


# Identification Pipeline
# Identification Pipeline
def identificationPipeline(img_path):
    '''
    Takes an image path, preprocesses it, and returns the predicted breed.
    '''
    print(f"Received img_path: {img_path}")

    # Validate Image Path
    if not os.path.exists(img_path):
        raise FileNotFoundError(f"Image not found at: {img_path}")

    # Read & Preprocess Image
    try:
        image = tf.io.read_file(img_path)
        image = tf.image.decode_jpeg(image, channels=3)
        image = tf.image.convert_image_dtype(image, tf.float32)
        image = tf.image.resize(image, size=[IMG_SIZE, IMG_SIZE])
        image_reshaped = tf.reshape(image, (1, IMG_SIZE, IMG_SIZE, 3))  # Add batch dimension

        # Predict
        results = loaded_full_model.predict(image_reshaped)
        confidence = np.max(results)
        pred_label = get_pred_label(results)
        
        if confidence < 0.1:
            return "Unknown Object (Not a Dog)"

        print(f"Predicted Label: {pred_label}")
    
        return pred_label
    
    except Exception as e:
        return f"Error: {str(e)}"