Package Summary

Tags No category tags.
Version 0.0.5
License MIT
Build type CATKIN
Use RECOMMENDED

Repository Summary

Checkout URI https://github.com/tue-robotics/image_recognition.git
VCS Type git
VCS Version master
Last Updated 2024-03-12
Dev Status DEVELOPED
CI status No Continuous Integration
Released RELEASED
Tags No category tags.
Contributing Help Wanted (0)
Good First Issues (0)
Pull Requests to Review (0)

Package Description

The image_recognition_rqt package

Additional Links

No additional links.

Maintainers

  • Rein Appeldoorn

Authors

No additional authors.

image_recognition_rqt

Contains rqt plugins for image recognition

Test plugin

RQT plugin to test servers that expose the interface defined in package: image_recognition_msgs

GUI Overview

How to

Start the rqt plugin stand alone (if this does not work, try a rqt --force-rediscover):

rosrun image_recognition_rqt test_gui # Calls a rqt -s image_recognition_rqt.test.TestPlugin

Select a rostopic of type sensor_msgs/Image and a service of type image_recognition_msgs/Recognize or image_recognition_msgs/GetFaceProperties with use of the configuration button in the menu-bar (gear wheel icon):

Select ROS topic

If you do not have any available topic, try to use your webcam with use of the usb_cam node (http://wiki.ros.org/usb_cam), you can start this node with:

rosrun usb_cam usb_cam_node # This step is optional ofcourse

Once you have an image stream in your GUI, you can draw a rectangle in the image feed, this ROI will be send to the object recognition srv.

See https://github.com/tue-robotics/image_recognition_openface for an example on faces.

Annotation plugin

GUI Overview

This RQT plugin (http://wiki.ros.org/rqt) enables easy labeling of objects using a graphical GUI. The end result will be a folder with label subdirs that contain image segments of the specified label, e.g. /tmp/object_recognition:

/tmp/object_recognition/:
drwxrwxr-x 2 rein rein 4,0K okt 18 21:22 apple
drwxrwxr-x 2 rein rein 4,0K okt 18 21:22 banana
drwxrwxr-x 2 rein rein 4,0K okt 18 21:22 choco_peanuts

/tmp/object_recognition/apple:
-rw-rw-r-- 1 rein rein 14K okt 18 21:22 2016-10-18-21-22-29.jpg
...

/tmp/object_recognition/banana:
-rw-rw-r-- 1 rein rein 16K okt 18 21:22 2016-10-18-21-22-37.jpg
...

/tmp/object_recognition/choco_peanuts:
-rw-rw-r-- 1 rein rein 8,6K okt 18 21:22 2016-10-18-21-22-43.jpg
...

This folder can be used for training for example a neural network. A Tensorflow example can be found here: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/examples/image_retraining/retrain.py

This plugin can also be used to test / use the Annotate.srv, see https://github.com/tue-robotics/image_recognition_openface

How to

Start the rqt plugin stand alone (if this does not work, try a rqt --force-rediscover):

rosrun image_recognition_rqt annotation_gui # Calls a rqt -s image_recognition_rqt.annotation.AnnotationPlugin

Select a rostopic of type sensor_msgs/Image with use of the configuration button in the menu-bar (gear wheel icon):

Select ROS topic

If you do not have any available topic, try to use your webcam with use of the usb_cam node (http://wiki.ros.org/usb_cam), you can start this node with:

rosrun usb_cam usb_cam_node # This step is optional ofcourse

You could also use the Folder image publisher plugin to publish images directly from a folder.

Once you have an image stream in your GUI, you can set the labels of your different objects with use of the 'Edit labels' button in the bottom-left corner:

Edit labels

It is also possible to specify your output path with use of the 'Edit path' button in the bottom-left corner:

Edit path

All these settings will be stored in the rqt config. Once you have set everything up, you can start labeling by drawing a rectangle in the image stream:

GUI Overview

A dialog will prompt that lets you select your object label and an image will be stored in the specified directory. If you want to capture more images for your object, just click the 'Save another one' button. If you want to save an image of a new object, redraw a rectangle and repeat the process.

Happy labeling :)

Manual plugin

Plugin for manually responding to a service call for classification

Folder image publisher

Plugin for publishing image from a folder via a topic. This tool can be used in combination with the annotation plugin.

CHANGELOG

Changelog for package image_recognition_rqt

0.0.5 (2019-06-08)

  • Added gender confidence to test tools
  • Added catkin lint to all packages
  • refactor package xml to 2.0
  • rename ROS pkgs with image_recognition_prefix
  • feat(rqt): Folder image publisher
  • feat(GetPersons.srv): Ability to call the GetPersons service
  • Contributors: Loy van Beek, Rein Appeldoorn

0.0.4 (2017-02-07)

0.0.3 (2017-02-07)

  • fix(): Add targets install for scripts
  • Contributors: Rein Appeldoorn

0.0.2 (2017-01-17)

  • fix(): Installation targets
  • Contributors: Rein Appeldoorn

0.0.1 (2017-01-17)

  • Removed unused function
  • Added GUI for manually recognizing from image by user
  • Added clear_on_click parameter in the constructor
  • Added method to obtain selected ROI
  • Added image recognition util to support labeled and raw writing of image in predefined folder structure
  • Update README.md
  • fixed ROIs
  • Added documentation to code
  • Updates to plugins to support the GetFaceProperties service
  • Working annotation gui and test gui for both image_recognition_tensorflow object recognition and image_recognition_openface face recognition
  • Bug in test gui
  • Tools working with new interface
  • Update README.md
  • Some minor reference and naming fixes
  • Merged object_recognition_rqt into image_recognition repo, renamed to image_recognition_rqt
  • Contributors: Rein Appeldoorn

Wiki Tutorials

See ROS Wiki Tutorials for more details.

Source Tutorials

Not currently indexed.

Launch files

No launch files found

Messages

No message files found.

Services

No service files found

Plugins

Recent questions tagged image_recognition_rqt at Robotics Stack Exchange