AI Image Search & Manager

A modern desktop application for indexing, searching, and managing large image collections using state-of-the-art Vision-Language Models (VLMs).

Concept

Managing thousands of images requires more than just file names. This application uses AI models like Qwen3-VL and Florence-2 to "look" at your photos, generate descriptive captions, and extract semantic features. This allows you to search your library by meaning (e.g., "a person wearing a red hat @sunset") rather than just manually assigned keywords.

Key Features

• Multi-Model Support: Switch between the powerful Qwen3-VL-2B (high quality) and the lightweight Florence-2-Base (high speed) depending on your hardware and needs.
• Intelligent Indexing:
  • Automated Captioning: Generates detailed descriptions for every image.
  • Smart Tagging: Uses NLP (spaCy) to extract relevant nouns and features as searchable tags.
  • Folder Synchronization: Scans for new, updated, or deleted images and keeps your database in sync.
• Advanced Search:
  • Semantic Search: Find images by describing their content in natural language.
  • Hybrid Search: Combine literal tag matching with semantic ranking for pinpoint accuracy.
• Custom Tag Management:
  • Manual Edits: Right-click any result to manually add, remove, or correct tags.
  • Global Manager: A central hub to browse and batch-edit metadata for all indexed images.
• Performance Optimized:
  • VRAM Aware: Automatically resizes images and clears GPU cache to prevent out-of-memory errors.
  • Async Search: Non-blocking search execution with visual progress feedback.
• Modern UI: Built with Flet (Flutter-based) featuring a sleek, responsive design with support for System/Light/Dark themes.

Installation

1. Clone the repository

git clone https://github.com/Yash-Kumar-tech/Image-Search.git
cd image-search

2. Set up environment

It is recommended to use a virtual environment (Conda or venv).

# Example with venv
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

Running the Application

Launch the application from the project root using:

python -m frontend.src.main

Usage

1. Initialize: The first launch will download the required AI models (weights are cached locally).
2. Index Folders: Click the Add Photo icon in the top bar, select a folder, and click "Index / Sync".
3. Search: Type anything in the search bar. The app uses Hybrid search by default to give you the most relevant results.
4. Manage Metadata:
   • Right-click search results to edit tags or open the file location.
   • Click the Tag icon in the top bar to open the central Metadata Manager.
5. Settings: Use the Settings icon to switch between Qwen-VL (Quality) and Florence-2 (Speed).

Technology Stack

• Frontend: Flet (Python-based Flutter wrapper)
• Backend AI: PyTorch, Hugging Face Transformers
• Databases:
  • SQLite: Image metadata and tags.
  • ChromaDB: High-dimensional vector embeddings for semantic search.
• NLP: spaCy (en_core_web_sm)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

---

Created for organizing extensive image repositories into structured, searchable knowledge bases.