Skip to content

Latest commit

 

History

History

image_recognition_tensorflow

README.md

Tensorflow image recognition

Object recognition with use of Tensorflow. Based on the retrain example: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/examples/image_retraining/retrain.py

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

Installation

See https://github.com/tue-robotics/image_recognition 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 https://github.com/tue-robotics/image_recognition/tree/master/image_recognition_rqt

  2. Retrain the neural network: https://github.com/tue-robotics/image_recognition/tree/master/image_recognition_tensorflow_rqt

  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]

  4. Test the classifier with use the test tool in https://github.com/tue-robotics/image_recognition/tree/master/image_recognition_rqt

ROS Node (object_recognition_node)

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

Scripts

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]