IR-Visible Dataset Preparation Tool

A desktop GUI application for aligning and cropping infrared (IR) and visible light image pairs to create synchronized datasets for multimodal computer vision tasks.

Features

• Automatic Pair Detection: Automatically matches IR (_T.JPG) and Visible (_Z.JPG) image pairs
• Interactive Alignment: Drag-and-drop alignment of visible images over IR base images
• Smart Cropping: Extract synchronized crops from both modalities simultaneously
• Progress Tracking: Resume work where you left off with JSON-based progress saving
• Bilingual Interface: English/Chinese toggle (press L key)
• Keyboard Shortcuts: Full keyboard control for efficient workflow
• Flexible Crop Sizes: 128×128, 256×256, or 512×512 pixel crops
• Zoom Controls: Adjust visible image zoom for precise alignment
• Opacity Control: Adjust IR image transparency for better visibility

Use Cases

• Creating datasets for IR-Visible image fusion models
• Preparing training data for thermal object detection
• Generating aligned pairs for multimodal deep learning
• Data preparation for surveillance, medical imaging, or remote sensing applications

Installation

Prerequisites

• Python 3.6 or higher
• OpenCV
• Pillow (PIL)
• NumPy

Install Dependencies

pip install opencv-python pillow numpy