app.py

import cv2
import numpy as np
import torch
import streamlit as st
from PIL import Image  # Import Image from PIL
from transformers import AutoModelForImageClassification, AutoFeatureExtractor

# Load AI Model from Hugging Face
model_name = "microsoft/resnet-50"
model = AutoModelForImageClassification.from_pretrained(model_name)
feature_extractor = AutoFeatureExtractor.from_pretrained(model_name)

def process_frame(frame):
    # Convert frame to tensor
    inputs = feature_extractor(images=[frame], return_tensors="pt")
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    return predictions

def main():
    st.title("Smart Mirror AI")
    
    # Option to run camera or upload image
    run = st.checkbox("Run Camera")
    uploaded_image = st.file_uploader("Upload Image", type=["jpg", "jpeg", "png"])
    
    if run:
        cap = cv2.VideoCapture(0)
        frame_placeholder = st.empty()
        
        while run:
            success, frame = cap.read()
            if not success:
                st.write("Failed to capture video")
                break
            else:
                frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
                predictions = process_frame(frame)
                frame_placeholder.image(frame, channels="RGB")
        cap.release()
    
    elif uploaded_image is not None:
        # If an image is uploaded, process it
        image = Image.open(uploaded_image)  # Use PIL.Image to open the image
        image = np.array(image)
        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
        predictions = process_frame(image)
        st.image(image, channels="BGR")
        st.write(f"Predictions: {predictions}")
    
if __name__ == "__main__":
    main()