Package Summary

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

Repository Summary

Checkout URI
VCS Type git
VCS Version master
Last Updated 2024-05-05
CI status No Continuous Integration
Tags No category tags.
Contributing Help Wanted (0)
Good First Issues (0)
Pull Requests to Review (0)

Package Description

The image_recognition_tensorflow package

Additional Links

No additional links.


  • Rein Appeldoorn


No additional authors.

Tensorflow image recognition

Object recognition with use of Tensorflow. Based on the retrain example:

RQT tools related to this package can be found in package image_recognition_tensorflow_rqt


See for installation instructions. Also make sure that you have a working installation on Tensorflow. If you are running Ubuntu x64, Tensorflow can be installed with use of the following commands:

sudo apt-get install python-pip python-dev
sudo pip install tensorflow # If you want gpu support, use pip install tensorflow-gpu

Quick How-to

  1. Annotate images with use of annotation tool provided in
  2. Retrain the neural network:
  3. Start the ROS node with the net as parameter:
    rosrun image_recognition_tensorflow object_recognition_node _graph_path:=[path_to_graph.pb] _labels_path:=[path_to_labels.txt]

  1. Test the classifier with use the test tool in

ROS Node (object_recognition_node)

rosrun image_recognition_tensorflow object_recognition_node _graph_path:=[path_to_graph.pb] _labels_path:=[path_to_labels.txt]


Get object recognition (get_object_recognition)

Get the classification result of an input image:

rosrun image_recognition_tensorflow get_object_recognition [path_to_graph.pb] [path_to_labels.txt] [path_to_image]

Evaluate classifier (evaluate_classifier)

Evaluate the classifier based on a structured images folder. The script assumes that the images in this directory are separated in different directories with the label name as directory name.

rosrun image_recognition_tensorflow evaluate_classifier [path_to_graph.pb] [path_to_labels.txt] [path_to_image_dir]

Retrain the neural network (retrain)

Train the neural network based on a set of images. The script assumes that the images in this directory are separated in different directories with the label name as directory name.

rosrun image_recognition_tensorflow retrain [image_folder] [model_folder_inceptionv3] [output_dir]


^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Changelog for package image_recognition_tensorflow ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

0.0.5 (2019-06-08)

  • Added test for image_recognition_tensorflow
  • Added catkin lint to all packages
  • refactor package xml to 2.0
  • rename ROS pkgs with image_recognition_prefix
  • Contributors: Matthijs van der Burgh, Rein Appeldoorn

0.0.4 (2017-02-07)

0.0.3 (2017-02-07)

  • fix(): Add targets install for scripts
  • Add numpy run dependency for image_recognition_tensorflow
  • Contributors: Matthijs van der Burgh, Rein Appeldoorn

0.0.2 (2017-01-17)

0.0.1 (2017-01-17)

  • Add try except if path already exists
  • Update retrain to the latest version of tensorflow
  • Added image recognition util to support labeled and raw writing of image in predefined folder structure
  • Update
  • Update
  • Documentation
  • Working annotation gui and test gui for both image_recognition_tensorflow object recognition and image_recognition_openface face recognition
  • Working with updated interface
  • Update
  • os.path.expanduser for paths
  • Some minor reference and naming fixes
  • Merged tensorflow_object_recognition into image_recognition repo, renamed to image_recognition_tensorflow
  • Contributors: Ramon Wijnands, Rein Appeldoorn

Wiki Tutorials

See ROS Wiki Tutorials for more details.

Source Tutorials

Not currently indexed.

Launch files

No launch files found


No message files found.


No service files found


No plugins found.

Recent questions tagged image_recognition_tensorflow at Robotics Stack Exchange