Home
>>> Russian Version (outdated)
This package is a solution in the field of computer vision, which contains a set of various methods for detecting objects, united by one interface. To use the package, you do not need to dive into the API, all detection settings are available through a single configuration file.
The package is developed within the "recognition by parts" paradigm, which allows you to specify various attributes of an object, such as image, color, shape, size, etc. These attributes are recognized separately and then aggregated into integral objects.
In addition to recognizing simple objects, the package allows you to specify their combinations, detecting a complex object, which is a set of simple objects and relationships between them.
The package provides a full-fledged ROS interface that allows you to use it in robotics tasks.
Some demos and intros to package, but unfortunatly exclusively on russian.
1.1. Nodes
1.2. Messages
1.3. Services
2.1. Common structure of the configuration file
2.2. Simple objects
2.3. Complex objects
3.1. Package install
3.2. Zbar (optional)
3.3. OpenCV contrib for ROS Noetic (optional)
3.4. OpenCV 4.2.0 for ROS Melodic and below (optional)(wip)
3.5. Libtorch (optional)(wip)
3.6. Dlib (optional)
3.7. igraph (optional)
3.8. Install on Jetson
4.1. Test launch (camera and kinect)
4.2. Test launch on bags
4.3. Supported videos
5.1. Error description
5.2. Feedback
6.1. TO DO list
6.2. Adding own attributes guide (wip)
6.3. Adding own relationships guide (wip)