ImageSorter — CLIP-Based Image Organizer with GUI

A desktop tool that uses OpenAI’s CLIP model to automatically classify and organize images into user-defined categories.
Images are assigned to one and only one category folder based on the highest-scoring CLIP similarity.

Includes a full Tkinter GUI for easy use.

---

✨ Features

• Organizes only images (.jpg, .jpeg, .png, .gif, .bmp, .tiff, .webp)

• Uses CLIP (openai/clip-vit-base-patch32) for semantic image understanding

• Images go to exactly one category

• Fully custom categories

  • Load from .txt
  • Or enter comma-separated list

• Dry-run mode for safe previews

• Move or Copy modes

• Progress bar + live console output

• Multi-threaded worker keeps UI responsive

• Saves CSV log (when not dry-run)

• Works on Windows / Linux / macOS

• Demo

---

📦 Installation

1. Clone the repository

git clone https://github.com/<your-user>/<your-repo>.git
cd <your-repo>

2. Create a virtual environment

python -m venv .venv

3. Activate the venv (PowerShell)

.\.venv\Scripts\Activate.ps1

If PowerShell blocks scripts:

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\.venv\Scripts\Activate.ps1

4. Install dependencies

pip install -r requirements.txt

---

▶️ Running the GUI

python organize_images_clip_gui.py

A window will appear with all available options.

---

🗂 Categories

You can define categories using either method:

Option A — categories.txt

Example:

People
Food
Screenshots
Documents
Pets
Landscapes
Receipts

Load it via Load from File in the GUI.

Option B — Inline (comma-separated)

People, Food, Screenshots, Pets

---

🛠 How It Works

1. Recursively scans the chosen root folder for images.

2. For each category, CLIP evaluates prompts like:

   "a photo of people"
   "a photo of food"
   "a photo of receipts"
   

3. Picks the highest-probability category if above threshold.

4. Moves/Copies the image into:

   <target>/<CategoryName>/
   

5. If no category scores high enough:

   <target>/Uncategorized/
   

---

📄 Log Output

When not in dry-run mode, saves:

<target>/organize_images_log.csv

Columns include:

• src
• dst
• category
• score

---

💻 Requirements

torch>=2.0.0
torchvision>=0.15.0
transformers>=4.30.0
pillow>=9.0.0
tqdm>=4.60.0
pandas>=1.3.0
typing-extensions

If you want GPU acceleration, install PyTorch with CUDA via: https://pytorch.org

---

📁 Folder Structure Example

/
├── organize_images_clip_gui.py
├── requirements.txt
├── categories.txt
└── README.md

---

🧪 Tips

• Enable dry-run first to verify.
• Use cuda device for faster sorting (if available).
• Good threshold values: 0.20–0.35.

---