app.py

import streamlit as st
from ultralytics import YOLO
import os
import zipfile

# Function to unzip the dataset (if not already extracted)
def unzip_dataset(zip_path, extract_path):
    if not os.path.exists(extract_path):
        os.makedirs(extract_path)
    with zipfile.ZipFile(zip_path, 'r') as zip_ref:
        zip_ref.extractall(extract_path)
    st.success("Dataset extracted successfully!")

# Paths to the dataset and model weights
zip_path = 'Clothing Segmentation DataSet.v9i.yolov8.zip'  # File should be uploaded in root
extract_path = '/tmp/clothing_dataset'  # /tmp is writable
model_weights = '/tmp/clothing_model.pt'  # Place to save/load model weights if needed


# Check if the model is already trained or needs to be trained
model_trained = os.path.exists(model_weights)

if not model_trained:
    # Unzip dataset if necessary
    unzip_dataset(zip_path, extract_path)

    # Train the model
    model = YOLO('yolov8s.pt')  # Using YOLOv8 small model for faster training

    # Path to the data.yaml
    data_yaml_path = os.path.join(extract_path, 'data.yaml')

    # Train the model
    with st.spinner('Training model...'):
        model.train(data=data_yaml_path, epochs=50, imgsz=640, batch=16)
        model.save(model_weights)  # Save the model after training
        st.success("Model trained and saved successfully!")

# Load the trained YOLOv8 model
model = YOLO(model_weights).to('cpu')

# Streamlit App Interface
st.title("Clothing Item Segmentation")

# Upload an image for segmentation
uploaded_image = st.file_uploader("Upload an image of clothing...", type=["jpg", "png", "jpeg"])

if uploaded_image is not None:
    # Display uploaded image
    st.image(uploaded_image, caption="Uploaded Image", use_column_width=True)

    # Run inference using the trained YOLOv8 model
    st.spinner("Running segmentation...")
    results = model.predict(source=uploaded_image)  # Run inference

    # Show the segmented image result
    st.image(results[0].plot(), caption="Segmentation Result", use_column_width=True)

    # Display the prediction details (class names and confidence)
    st.write("Prediction Details:")
    for result in results[0].boxes.data:
        class_id = int(result[5])  # Class ID
        confidence = result[4].item()  # Confidence
        st.write(f"Class ID: {class_id}, Confidence: {confidence:.2f}")