https://github.com/opencv/open_vision_capsules
A set of libraries for encapsulating smart vision algorithms
https://github.com/opencv/open_vision_capsules
Last synced: about 1 year ago
JSON representation
A set of libraries for encapsulating smart vision algorithms
- Host: GitHub
- URL: https://github.com/opencv/open_vision_capsules
- Owner: opencv
- License: other
- Created: 2020-01-17T01:29:06.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2025-01-08T21:30:27.000Z (over 1 year ago)
- Last Synced: 2025-03-28T13:08:46.047Z (over 1 year ago)
- Language: Python
- Size: 549 KB
- Stars: 134
- Watchers: 12
- Forks: 48
- Open Issues: 7
-
Metadata Files:
- Readme: README.md
- License: COPYING
Awesome Lists containing this project
README
# OpenVisionCapsules
[](https://openvisioncapsules.readthedocs.io/en/latest/?badge=latest)
This repository contains the OpenVisionCapsules SDK, a set of Python libraries
for encapsulating machine learning and computer vision algorithms for
intelligent video analytics.
Encapsulating an algorithm allows it to be deployed as a single, self-describing
file that inputs and outputs data in a standard format. This makes deployment
and integration significantly easier than starting with a model file or a
snippet of source code. Capsules are descriptive of their input and output
requirements, allowing OpenVisionCapsules to route data between capsules
automatically.
This project is split into two packages, `vcap` and `vcap-utils`. `vcap`
contains the necessary facilities to create and encapsulate an algorithm.
`vcap-utils` contains a set of utilities that make encapsulating algorithms of
certain types easier.
# Project Status
OpenVisionCapsules is in a developer preview phase. We're looking for developer
feedback before reaching a stable 1.0 release. If you find any bugs or have
suggestions, please open an issue.
# Getting Started
Take a look at the [documentation here][docs].
A couple example capsules are available under `vcap/examples`, demonstrating
how to create classifier and detector capsules from TensorFlow models.
# Installation
To install OpenVisionCapsules locally, clone the repository and run the
following commands to install the `vcap` and `vcap-utils` packages in the
current environment.
```
pip3 install -e ./vcap
pip3 install -e ./vcap_utils
```
# Examples
To make use of the example capsules in the `vcap/examples/` directory, make
sure to run the tests with pytest (from the root of the repo). The tests
download all the necessary models and images, including the models for the
example capsules.
Make sure vcap & vcap-utils installation is done before tests,
```
pytest -v -x .
```
A repository of open source capsules can be found [here][capsule_zoo].
[docs]: https://openvisioncapsules.readthedocs.io/en/latest/
[capsule_zoo]: https://github.com/aotuai/capsule_zoo
capsule inference test,
```
python3 tools/openvisioncapsule_tools/capsule_infer/capsule_infer.py --capsule ../capsule-zoo/capsules/detector_person_vehicle_bike_openvino --images tests/test_images/two_people.jpg
```