box-diff
Utilities for comparing bounding boxes
Author: Kevin Barnard, kbarnard@mbari.org
Installation
box-diff is available on PyPI as boxdiff:
pip install boxdiff
Usage
All of box-diff is housed within the boxdiff package.
import boxdiff
The core data models are defined in the boxdiff.models module. These are split into three groups:
BoundingBox: ID + labeled 2D bounding boxImage: ID + collection of bounding boxesImageSet: ID + collection of images
Note: IDs may be integers, UUIDs, or strings.
Each group has a data model (defined in boxdiff.models.core), a delta (boxdiff.models.deltas), and a difference flag (boxdiff.models.flags).
The data model represents the object and its attributes, whereas a delta represents the difference in attribute values between two objects.
Difference flags represent the presence of attribute differences between two objects as derived from a delta object.
All of the data models are serializable to JSON. For example,
json_str = bounding_box.to_json(indent=2)
print(json_str)
might give
{
"id": 0,
"label": "label",
"x": 0.0,
"y": 0.0,
"width": 1.0,
"height": 1.0
}
Similarly, data model objects may be parsed from JSON. For example,
bounding_box = BoundingBox.from_json(json_str)
print(bounding_box)
# BoundingBox(id=0, label='label', x=0.0, y=0.0, width=1.0, height=1.0)
Bounding Boxes
A BoundingBox is defined by an ID, a label, and a 2D box (x, y, width, height).
from boxdiff import BoundingBox
car_box = BoundingBox(
id=0,
label='car',
x=100, y=200,
width=300, height=80
)
Equality can be checked using the == operator:
same_car_box = BoundingBox(
id=0,
label='car',
x=100, y=200,
width=300, height=80
)
print(car_box == same_car_box)
# True
corrected_car_box = BoundingBox(
id=0,
label='car',
x=90, y=210,
width=320, height=85
)
print(car_box == corrected_car_box)
# False
A BoundingBoxDelta between two boxes can be computed with the - operator:
box_delta = corrected_car_box - car_box
print(box_delta)
# BoundingBoxDelta(id=1, label_old='car', label_new='car', x_delta=-10.0, y_delta=10.0, width_delta=20.0, height_delta=5.0)
BoundingBoxDifference flags may then be computed from the delta:
print(box_delta.flags)
# BoundingBoxDifference.RESIZED|MOVED
Deltas may be applied to a bounding box using the + operator:
new_car_box = car_box + box_delta
print(new_car_box)
# BoundingBox(id=1, label='car', x=90.0, y=210.0, width=320.0, height=85.0)
print(new_car_box == corrected_car_box)
# True
Area may be computed from a bounding box:
print(car_box.area)
# 24000.0
Intersection over union between bounding boxes can be computed using the iou method:
car_iou = car_box.iou(corrected_car_box)
print(car_iou)
# 0.695364238410596
Images
An Image is defined by an ID and a collection of bounding boxes.
from boxdiff import Image
from uuid import UUID
image = Image(
id=UUID('78d76772-4664-467c-ae88-a25496234966'),
bounding_boxes=[car_box]
)
Likewise, equality can be checked using the == operator and deltas computed using the - operator:
corrected_image = Image(
id=UUID('78d76772-4664-467c-ae88-a25496234966'),
bounding_boxes=[corrected_car_box]
)
image_delta = corrected_image - image
print(image_delta)
# ImageDelta(
# id=0,
# boxes_added=[],
# boxes_removed=[],
# box_deltas=[
# BoundingBoxDelta(
# id=0,
# label_old='car',
# label_new='car',
# x_delta=-10.0,
# y_delta=10.0,
# width_delta=20.0,
# height_delta=5.0
# )
# ]
# )
Similarly, flags may be computed from the delta:
print(image_delta.flags)
# ImageDifference.BOXES_MODIFIED
Image Sets
An ImageSet is defined by an ID and a collection of images.
from boxdiff import ImageSet
image_set = ImageSet(
id='my_image_set',
images=[image, ...]
)
Its syntax and structure is analogous to that of an Image.
Metadata
Release files for boxdiff 0.2.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| boxdiff-0.2.0.tar.gz | 4.4 kB | Without provenance | Details for boxdiff-0.2.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| boxdiff-0.2.0-py3-none-any.whl | Python 3 | none | any | Details for boxdiff-0.2.0-py3-none-any.whl |
Total release size: 10.0 kB
Release files / boxdiff-0.2.0.tar.gz
| Download URL | boxdiff-0.2.0.tar.gz |
|---|---|
| Size | 4.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
02f1fbead0f9710b4260b71ed69be6b31d066c614efff439b0af6697ce7fef1a
|
|
BLAKE2b-256 checksum How to use checksums |
5e4e3946ae2c95d81034513d8c4e9dde55abb59e0ac0d9770e9d04c9069d368c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
poetry/1.6.1 CPython/3.11.4 Linux/6.5.0-15-generic
|
Release files / boxdiff-0.2.0-py3-none-any.whl
| Download URL | boxdiff-0.2.0-py3-none-any.whl |
|---|---|
| Size | 5.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
e771da66a7b6c5cbeb7582a5e35eb0a43b814415d86b62fd5aa35c066f40f7ed
|
|
BLAKE2b-256 checksum How to use checksums |
dcbeacbd8b41e0d45cffe37ad341d9ef99464f61018a2aae41f61d8e4db00ab0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
poetry/1.6.1 CPython/3.11.4 Linux/6.5.0-15-generic
|
Security for boxdiff 0.2.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.