bbox-visualizer 1.0.0

Different ways of visualizing objects given bounding box data

bbox-visualizer

This package helps users draw bounding boxes around objects, without doing the clumsy math that you'd need to do for positioning the labels. It also has a few different types of visualizations you can use for labeling objects after identifying them.

By default the bounding box points are expected in Pascal VOC format: (xmin, ymin, xmax, ymax). COCO and YOLO formats are also supported via the bbox_format keyword argument (see Bounding box formats).

• Documentation: https://bbox-visualizer.readthedocs.io.
• Free software: MIT license

Installation:

pip install bbox-visualizer

Quick Start

A complete example that loads an image, draws a labeled bounding box, and saves the result:

import cv2
import bbox_visualizer as bbv

img = cv2.imread("path/to/image.jpg")

# Bounding boxes use [x_min, y_min, x_max, y_max]
bbox = [150, 100, 450, 300]
label = "person"

img = bbv.draw_box(img, bbox, bbox_color=(0, 255, 0))
img = bbv.add_label(img, label, bbox)

cv2.imwrite("output.jpg", img)

All functions return a new image and never modify the input image, so keep the return value (as above) rather than relying on in-place changes.

For multiple objects, use the _multiple_ variants with parallel lists:

bboxes = [[150, 100, 450, 300], [500, 50, 700, 250]]
labels = ["person", "dog"]

img = bbv.draw_multiple_boxes(img, bboxes)
img = bbv.add_multiple_labels(img, labels, bboxes)

draw_multiple_boxes also accepts one color per box:

img = bbv.draw_multiple_boxes(img, bboxes, bbox_color=[(0, 255, 0), (0, 0, 255)])

The library logs fallback warnings (e.g., when a label doesn't fit) through Python's logging module. To silence them:

import logging

logging.getLogger("bbox_visualizer").setLevel(logging.ERROR)

Bounding box formats

Every drawing function accepts a bbox_format keyword argument. The default is Pascal VOC.

bbox_format	Coordinates	Scale
"voc" (default)	[x_min, y_min, x_max, y_max]	absolute pixels
"coco"	[x_min, y_min, width, height]	absolute pixels
"yolo"	[x_center, y_center, width, height]	normalized to [0, 1]

# COCO format: [x_min, y_min, width, height]
img = bbv.draw_box(img, [150, 100, 300, 200], bbox_format="coco")

# YOLO format: [x_center, y_center, width, height], normalized to [0, 1].
# Image dimensions are read from the image, so no extra arguments are needed.
img = bbv.draw_box(img, [0.5, 0.4, 0.3, 0.25], bbox_format="yolo")

# Works with the multiple-object variants too
img = bbv.draw_multiple_boxes(img, coco_bboxes, bbox_format="coco")

Internally all formats are converted to Pascal VOC before drawing.

Runnable scripts live in examples/:

• quickstart.py — minimal example on a blank canvas
• single_object.py — every single-object label style
• multiple_objects.py — every multi-object label style
• label_stress.py — awkward label strings through every style, for eyeballing label layout

Photos by Joshua Earle, Jonas Weckschmied and Sherzod Max on Unsplash.

image	function
	img = bbv.draw_box(img, bbox) img = bbv.add_label(img, label, bbox, top=True)
	img = bbv.draw_box(img, bbox) img = bbv.add_T_label(img, label, bbox)
	img = bbv.draw_flag_with_label(img, label, bbox)
	img = bbv.draw_box(img, bbox) img = bbv.add_label(img, label, bbox, top=False)
	img = bbv.draw_box(image, bbox, is_opaque=True) img = bbv.add_label(img, label, bbox, draw_bg=False, top=False)
	img = bbv.draw_multiple_boxes(img, bboxes) img = bbv.add_multiple_labels(img, labels, bboxes)
	img = bbv.draw_multiple_flags_with_labels(img, labels, bboxes)
	img = bbv.draw_multiple_boxes(img, bboxes) img = bbv.add_multiple_T_labels(img, labels, bboxes)

Note: The functions draw_rectangle and draw_multiple_rectangles are also available as aliases for draw_box and draw_multiple_boxes respectively. Both naming conventions work identically.

Tip: The draw_multiple_* and add_multiple_* functions are convenience helpers. For full control over your visualizations, call the single-box functions (draw_box, add_label, etc.) in a loop instead.