A web-based tool for managing and reviewing Qwen training datasets with image comparison features.
- 📁 Folder Selection - Browse and select dataset folders
- 🖼️ Image Grid - View all images from the dataset in a responsive grid
- 🔍 Fullscreen Preview - Click any image to view in fullscreen
- 🎨 Overlay Comparison - Toggle between normal view and semi-transparent overlay to compare with Control1 images
- ⌨️ Keyboard Navigation - Navigate with arrow keys, toggle with space, delete with backspace
- 🗑️ Batch Deletion - Delete all related images (img, Control1, Control2) at once
Your dataset folder should have this structure:
DatasetFolder/
├── img/ # Source images
├── Control1/ # First control images
└── Control2/ # Second control images
All three folders must contain images with matching filenames (e.g., image_00003_.png).
-
Install dependencies:
pip install -r requirements.txt
-
Run the server:
python app.py
-
Open your browser: Navigate to
http://localhost:5000
- Select a dataset folder from the dropdown menu
- Browse images in the grid view
- Click an image to open fullscreen preview
- Toggle overlay to compare img with Control1 (img becomes semi-transparent)
- Navigate using arrow keys or on-screen buttons
- Delete mismatched sets by pressing Backspace/Delete
| Key | Action |
|---|---|
← / → |
Navigate between images |
Space |
Toggle overlay |
Backspace / Delete |
Delete current image set (all 3 files) |
Esc |
Close preview |
- Backend: Python Flask
- Frontend: Vanilla JavaScript, HTML5, CSS3
- Design: Modern dark theme with glassmorphism and smooth animations
MIT