shoumikchow/bbox-visualizer

What it solves

It simplifies the process of drawing bounding boxes and labels on images, removing the need for manual coordinate calculations for label positioning. It provides various visualization styles to help users identify and label objects in computer vision tasks.

How it works

The library provides functions to draw boxes and labels on images using OpenCV. It supports three common bounding box formats: Pascal VOC (absolute pixels), COCO (absolute pixels), and YOLO (normalized coordinates). Internally, it converts all formats to Pascal VOC before rendering. All functions return a new image rather than modifying the original input image.

Who it’s for

Developers and researchers working with object detection models and computer vision projects who need a way to quickly visualize their model's predictions or ground truth labels.

Highlights

• Supports multiple bounding box formats: VOC, COCO, and YOLO.
• Offers various label styles, including labels on top, inside the box, T-labels, and flag-style labels.
• Provides convenience functions for drawing multiple boxes and labels simultaneously.
• Ensures non-destructive editing by returning new image objects.