pybboxes

What it is and what it does

PyBboxes is a lightweight toolkit for working with bounding boxes in computer vision workflows. It handles conversion between five common bounding box formats used across popular frameworks and datasets: COCO (top-left + width/height), YOLO (center + normalized width/height), VOC (top-left + bottom-right), FiftyOne (normalized COCO), and Albumentations (normalized VOC). The package also provides geometric operations like IoU computation, area calculation, and intersection/union calculations between boxes.

The library is built on numpy, pycocotools, and pyyaml and supports both strict and lenient modes for handling out-of-bounds boxes. It includes an annotation file converter for batch processing YOLO, COCO, and VOC format files. Typical use cases are converting between dataset formats during preprocessing, computing overlap metrics during model evaluation, and batch-converting annotation files across different computer vision frameworks.

Use it for:

• Convert bounding boxes between COCO and YOLO formats when switching between dataset sources or training frameworks
• Compute IoU and intersection metrics for non-maximum suppression or model evaluation pipelines
• Batch-convert annotation files across YOLO, COCO, and VOC formats during dataset preparation
• Validate and filter out-of-bounds boxes with strict mode during data cleaning
• Perform geometric operations (area, union, difference) on detected objects for post-processing

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Converts between bounding box formats (COCO, YOLO, VOC, FiftyOne, Albumentations) and computes geometric operations like IoU, area, and intersection.

Yes. PyBboxes is a focused, low-friction utility with no known vulnerabilities, permissive MIT licensing, and stable production status. It solves a concrete problem in computer vision workflows—format conversion and box geometry—that most practitioners encounter. Dormant maintenance is acceptable for a mature, feature-complete library; the last commit (2024-10-07) is recent enough to indicate active monitoring. Install if you work with multiple bounding box formats or need reliable geometric operations.

Install

pybboxes on PyPI

pip install pybboxes

uv add pybboxes

poetry add pybboxes

Installing pybboxes

Before you install

Low friction install with three stable dependencies (numpy, pycocotools, pyyaml). Maintenance is dormant—last commit was 2024-10-07 with no activity since—but the package is marked Production/Stable and has seen no breaking changes in recent releases.

License in practice

MIT license permits commercial and private use with minimal restrictions; you must include a copy of the license in distributions.

Quickstart

pip install pybboxes

from pybboxes import BoundingBox

my_coco_box = [98, 345, 322, 117]
coco_bbox = BoundingBox.from_coco(*my_coco_box, image_size=(640, 480))
voc_bbox = coco_bbox.to_voc()

Requires Python >= 3.8. Conversion between normalized formats (YOLO, FiftyOne, Albumentations) requires image_size to be set.

Verify before relying

• Whether pycocotools is a hard requirement or optional for specific use cases
• Performance characteristics when handling large batches of bounding boxes
• Whether annotation file conversion supports all three formats (YOLO, COCO, VOC) bidirectionally

Package facts

Evidence: pybboxes-0.2.0-py3-none-any.whl

Keywords: machine-learning, deep-learning, image-processing, pytorch, tensorflow, numpy, bounding-box, iou, computer-vision, cv

Tags

More Libraries packages