app.py

import streamlit as st
import tensorflow as tf
import numpy as np
from PIL import Image
import os
from huggingface_hub import login
from tensorflow import keras
import requests

# Get Hugging Face token from environment variables
HF_TOKEN = os.environ.get("HF_TOKEN")

if not HF_TOKEN:
    st.error("HF_TOKEN environment variable not set. Please set it before running the app.")
    st.stop()  # Stop execution if token is missing

# Authenticate with Hugging Face
try:
    login(token=HF_TOKEN)
    st.success("Successfully logged in to Hugging Face Hub!")
except Exception as e:
    st.error(f"Hugging Face login failed: {e}")
    st.stop()

model_url = "https://huggingface.co/louiecerv/cats_dogs_recognition_tf_cnn/resolve/main/cats_dogs_recognition_tf_cnn.keras"
model_filename = "cats_dogs_recognition_tf_cnn.keras"  # Or any name you prefer
model_filepath = model_filename  # Save in the current directory

try:
    # 1. Download the model file
    response = requests.get(model_url, stream=True)
    response.raise_for_status()  # Raise an exception for bad status codes (4xx or 5xx)
    with open(model_filepath, "wb") as f:
        for chunk in response.iter_content(chunk_size=8192):
            f.write(chunk)

    # 2. Load the model from the local file
    model = tf.keras.models.load_model(model_filepath)
    st.success(f"Model loaded successfully from: {model_filepath}")

    # Optional: Clean up the downloaded file (if desired)
    # os.remove(model_filepath)  # Uncomment if you want to delete the file after loading

except requests.exceptions.RequestException as e:
    st.error(f"Error downloading model: {e}")
    st.stop()
except Exception as e:
    st.error(f"Error loading model: {e}")
    st.stop()

# Image preprocessing function with cropping
def preprocess_image(image):
    # 1. Crop to square aspect ratio
    width, height = image.size
    new_size = min(width, height)
    left = (width - new_size) // 2
    top = (height - new_size) // 2
    right = (width + new_size) // 2
    bottom = (height + new_size) // 2
    image = image.crop((left, top, right, bottom))

    # 2. Resize
    image = image.resize((128, 128))  # Use PIL's resize for consistency

    # 3. Convert to NumPy array if it's a PIL image
    image_np = np.array(image)

    # 4. Ensure 3 channels (RGB)
    if image_np.ndim == 2 or image_np.shape[2] == 1:
        image_np = np.stack([image_np] * 3, axis=-1)

    # 5. Normalize
    image_np = image_np.astype(np.float32) / 255.0
    image_np = np.expand_dims(image_np, axis=0)
    return image_np

# Streamlit app
st.title("Cat vs. Dog Image Classifier")

uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "png", "jpeg"])

if uploaded_file is not None:
    image = Image.open(uploaded_file)
    st.image(image, caption="Uploaded Image.", use_container_width=True)

    processed_image = preprocess_image(image)  # Pass the PIL Image directly

    # Display preprocessed image (optional, but helpful for debugging)
    processed_image_display = tf.squeeze(processed_image, axis=0).numpy() * 255.0
    processed_image_display = processed_image_display.astype(np.uint8)
    processed_image_display = Image.fromarray(processed_image_display)
    st.image(processed_image_display, caption="Preprocessed Image", use_container_width=True)


    predictions = model.predict(processed_image)
    class_index = np.argmax(predictions)
    class_names = ["Cat", "Dog"]
    predicted_class = class_names[class_index]
    confidence = predictions[0][class_index] * 100

    st.write(f"## Prediction: {predicted_class} (Confidence: {confidence:.2f}%)")