AI-Powered Interactive Object Color Visualization System

Overview

This application allows users to upload interior images, select objects (walls, furniture) via AI-assisted segmentation (SAM), and recolor them realistically using LAB color space blending.

Prerequisites

• Python 3.8+
• CUDA-capable GPU (Recommended) or CPU

Installation

1. Clone/Open Project Directory

   cd "Python Image with color"

2. Install Dependencies

   pip install -r requirements.txt

3. Download Model Weights If the weights/ folder is empty, run:

   python download_weights.py

   This downloads the sam_vit_b_01ec64.pth checkpoint.

Running the Application

Run the Streamlit app:

streamlit run app.py

Usage Guide

1. Upload: Drag and drop an interior image (JPG/PNG).
2. Wait: The system will compute embeddings (takes ~5-10s on CPU initially).
3. Select: Click on the object you want to recolor on the image.
4. Recolor: Pick a color from the sidebar and click "Apply Color".
5. Iterate: Use "Undo" to revert changes.
6. Download: Click "Download Image" to save your result.

Architecture

• Frontend: Streamlit
• AI Model: Segment Anything Model (ViT-B)
• Color Engine: OpenCV (LAB Color Space)

Deployment (Render)

This project includes configuration for deploying to Render.

1. Push this code to a GitHub or GitLab repository.
2. Log in to Render and click New + -> Web Service.
3. Connect your repository.
4. Render should automatically detect the render.yaml file.
   • If manually configuring:
     • Runtime: Python 3
     • Build Command: ./build.sh
     • Start Command: streamlit run app.py --server.port $PORT --server.address 0.0.0.0
5. Important: Select the Starter plan or higher. The AI model requires significant RAM and may crash on the Free tier.