kvnptl / mas_perception_libs

mas_perception_libs

Library containing shared perception functionality. Python interface are developed for several functions and definitions in the C++ library using Boost. This is to avoid defining commonly used logic in different locations and also to expose C++ exclusive libraries (i.e. PCL) for use in Python. This means if a functionality is needed in both Python and C++, it should first be developed in C++ and then ported to Python.

• mas_perception_libs
  • Build issues
  • C++ Library
  • Python Package
  • Executables
    • object_detection_action_server
    • plane_detection_action_server
    • image_detection_test
    • image_recognition_server
    • image_recognition_client_test
    • cloud_processing_python_test
    • cloud_processing_cpp_test
  • Launch Files
    • plane_detection.launch
    • object_detection.launch
    • image_detection_test.launch
    • image_recognition.launch
    • cloud_processing_python_test.launch
    • cloud_processing_cpp_test.launch
  • Directory structure

Build issues

• numpy installation may not create a correct symbolic link to the library's include directory in the system, which results in a compilation error about missing header file numpy/arrayobject.h. This can be fixed by creating a correct link as follow (this is typical for an Ubuntu machine, please make sure it's appropriate for your own setup):

  # ln -s /usr/local/lib/python2.7/dist-packages/numpy/core/include/numpy /usr/include/numpy
  

C++ Library

Contains shared perception definitions in C++.

Python Package

Contains shared perception definitions in Python.

Executables

object_detection_action_server

Run an instance of ObjectDetectionActionServer (described in Python Documentation).

Parameters:

• action_name: name of action server
• cloud_topic: name of topic which supply sensor_msgs/PointCloud2 messages
• target_frame: name of reference frame which the object poses will be transformed to
• class_annotations: YAML file which maps numeric class values to class names, used by the ImageDetectorBase class to configure the detection model. An example of this file is class_annotation_example.yml.
• kwargs_file: YAML file which is used by the ImageDetectorBase class to specify additional parameters needed to configure the detection model. An example is image_detector_test_kwargs.yml.
• detection_module: name of the module containing the ImageDetectorBase extension to import.
• detection_class: name of the extension of the ImageDetectorBase class to import.

plane_detection_action_server

Run an instance of PlaneDetectionActionServer (described in Python Documentation).

Parameters:

• action_name: name of action server
• cloud_topic: name of topic which supply sensor_msgs/PointCloud2 messages
• target_frame: name of reference frame which the object poses will be transformed to
• class_annotations: YAML file which maps numeric class values to class names, used by the ImageDetectorBase class to configure the detection model. An example of this file is class_annotation_example.yml.
• kwargs_file: YAML file which is used by the ImageDetectorBase class to specify additional parameters needed to configure the detection model. An example is image_detector_test_kwargs.yml.
• detection_module: name of the module containing the ImageDetectorBase extension to import.
• detection_class: name of the extension of the ImageDetectorBase class to import.
• also parameters defined in the dynamic reconfiguration file PlaneFitting.cfg.

image_detection_test

Node for testing image detection models. Can test images from a directory, a sensor_msgs/Image topic, or a sensor_msgs/PointCloud2 topic.

Parameters:

• class_annotations, kwargs_file, detection_module, and detection_class: parameters for ImageDetectorBase class similar to ones described above for object_detection_action_server and plane_detection_action_server.
• result_topic: sensor_msgs/Image topic which visualized detection results are published.
• image_directory: if specified will ignore other image sources and read images from this directory for testing.
• cloud_topic: if specified and image_directory is not specified will extract images from sensor_msgs/PointCloud2 messages from this topic for testing.
• image_topic: if specified and the other 2 image sources are not specified will test detection model on sensor_msgs/Image messages from this topic.

image_recognition_server

Server which uses an instance of ImageClassifier class (See documentation) to classify images.

Parameters:

• service_name: name of recognition service to advertise (default: '~recognize_image').
• model_directory: directory containing the trained classification model (default: '')
• recognition_module: module containing the ImageClassifier instance (default: 'mas_perception_libs')
• recognition_class: class name of the ImageClassifier instance (default: 'ImageClassifierTest')

image_recognition_client_test

Script to test the image_recognition_server.

usage: image_recognition_client_test [-h] --test-dir TEST_DIR --service-name
                                     SERVICE_NAME [--num-samples NUM_SAMPLES]
                                     [--preprocess-input-module PREPROCESS_INPUT_MODULE]
                                     model_name

Tool to test model with test images using KerasImageClassifier class.Assuming
images to be of type jpg

positional arguments:
  model_name            Keras model to be tested

optional arguments:
  -h, --help            show this help message and exit
  --test-dir TEST_DIR, -t TEST_DIR
                        directory with test images
  --service-name SERVICE_NAME, -s SERVICE_NAME
                        name of recognition service
  --num-samples NUM_SAMPLES, -n NUM_SAMPLES
                        number of samples to test, if left blank, take all
                        samples.
  --preprocess-input-module PREPROCESS_INPUT_MODULE, -p PREPROCESS_INPUT_MODULE
                        module containing image preprocessing function.

Example execution:

image_recognition_client_test -s <service_name> -t <folder_with_test_images> <model_name>

cloud_processing_python_test

Script for testing PlaneSegmenter (see Python documentation), including handling of the dynamic reconfiguration defined in PlaneFitting.cfg. Parameters:

• cloud_topic (string): input point cloud topic
• processed_cloud_topic (string): topic on which the filtered cloud will be published
• target_frame (string): coordinate frame that the point cloud will be transformed to.
• extract_planes(bool): if false will only do cloud filtering
• parameters defined in PlaneFitting.cfg

cloud_processing_cpp_test

C++ executable testing PlaneSegmenterROS (see C++ documentation), including handling of the dynamic reconfiguration defined in PlaneFitting.cfg. Parameters:

• cloud_topic (string): input point cloud topic
• processed_cloud_topic (string): topic on which the filtered cloud will be published
• target_frame (string): coordinate frame that the point cloud will be transformed to.
• extract_planes(bool): if false will only do cloud filtering
• parameters defined in PlaneFitting.cfg

Launch Files

plane_detection.launch

Launch the plane_detection_action_server script. Arguments are the same with the script's parameters:

• action_name (default: "/mas_perception/detect_image")
• cloud_topic (default: "")
• target_frame (default: "/base_link")
• class_annotations (default: "$(find mas_perception_libs)/models/class_annotation_example.yml")
• kwargs_file (default: "$(find mas_perception_libs)/models/image_detector_test_kwargs.yml")
• detection_module (default: "mas_perception_libs")
• detection_class (default: "ImageDetectorTest")
• plane_fitting_config_file (default "$(find mas_perception_libs)/ros/config/plane_fitting_default_configs.yaml")

object_detection.launch

Launch the object_detection_action_server script. Arguments are the same with the script's parameters:

• action_name (default: "/mas_perception/detect_image")
• cloud_topic (default: "")
• target_frame (default: "/base_link")
• class_annotations (default: "$(find mas_perception_libs)/models/class_annotation_example.yml")
• kwargs_file (default: "$(find mas_perception_libs)/models/image_detector_test_kwargs.yml")
• detection_module (default: "mas_perception_libs")
• detection_class (default: "ImageDetectorTest")

image_detection_test.launch

Launch the image_detection_test script. Arguments are the same with the script's parameters:

• result_topic (default: "/mas_perception/detection_result")
• image_directory (default: "")
• cloud_topic (default: "")
• image_topic (default: "")
• class_annotations (default: "$(find mas_perception_libs)/models/class_annotation_example.yml")
• kwargs_file (default: "$(find mas_perception_libs)/models/image_detector_test_kwargs.yml")
• detection_module (default: "mas_perception_libs")
• detection_class (default: "ImageDetectorTest")

image_recognition.launch

Launch the image_recognition_server. Arguments:

• service_name: name of recognition service to advertise (default: '~recognize_image').
• model_directory: directory containing the trained classification model (default: $(find mas_perception_libs)/models)
• recognition_module: module containing the ImageClassifier instance (default: 'mas_perception_libs')
• recognition_class: class name of the ImageClassifier instance (default: 'ImageClassifierTest')

cloud_processing_python_test.launch

Launch the cloud_processing_python_test executable. Arguments are similar to ones for the executable, with an additional plane_fitting_config_file in which the default values for the dynamic reconfiguration PlaneFitting.cfg. Default file is plane_fitting_default_configs.yaml.

cloud_processing_cpp_test.launch

Launch the cloud_processing_cpp_test executable. Arguments are similar to ones for the executable, with an additional plane_fitting_config_file in which the default values for the dynamic reconfiguration PlaneFitting.cfg. Default file is plane_fitting_default_configs.yaml.

Directory structure

.
├── CMakeLists.txt
├── common
│   ├── include
│   │   └── mas_perception_libs
│   │       ├── aliases.h
│   │       ├── bounding_box_2d.h
│   │       ├── bounding_box.h
│   │       ├── impl
│   │       │   └── pyboostcvconverter.hpp
│   │       ├── point_cloud_utils.h
│   │       ├── sac_plane_segmenter.h
│   │       └── use_numpy.h
│   └── src
│       ├── bounding_box_2d.cpp
│       ├── bounding_box.cpp
│       ├── init_numpy_api.cpp
│       ├── point_cloud_utils.cpp
│       ├── pyboost_cv3_converter.cpp
│       └── sac_plane_segmenter.cpp
├── docs
│   ├── cpp_library.md
│   ├── image_detection_class_structure.png
│   └── python_package.md
├── models
│   ├── class_annotation_example.yml
│   ├── image_detector_test_kwargs.yml
│   └── test_model.txt
├── package.xml
├── README.md
├── ros
│   ├── config
│   │   ├── PlaneFitting.cfg
│   │   └── plane_fitting_default_configs.yaml
│   ├── include
│   │   └── mas_perception_libs
│   │       ├── bounding_box_wrapper.h
│   │       ├── color.h
│   │       ├── image_bounding_box.h
│   │       ├── impl
│   │       │   └── ros_message_serialization.hpp
│   │       └── point_cloud_utils_ros.h
│   ├── launch
│   │   ├── cloud_processing_cpp_test.launch
│   │   ├── cloud_processing_python_test.launch
│   │   ├── image_detection.launch
│   │   ├── object_detection.launch
│   │   ├── image_detection_test.launch
│   │   └── image_recognition.launch
│   ├── scripts
│   │   ├── cloud_processing_python_test
│   │   ├── plane_detection_action_server
│   │   ├── object_detection_action_server
│   │   ├── image_detection_test
│   │   ├── image_recognition_client_test
│   │   └── image_recognition_server
│   └── src
│       ├── boost_python_module.cpp
│       ├── bounding_box_wrapper.cpp
│       ├── cloud_processing_test_node.cpp
│       ├── image_bounding_box.cpp
│       ├── mas_perception_libs
│       │   ├── bounding_box.py
│       │   ├── constants.py
│       │   ├── image_classifier.py
│       │   ├── image_detector.py
│       │   ├── image_recognition_service.py
│       │   ├── __init__.py
│       │   ├── object_detector.py
│       │   ├── ros_message_serialization.py
│       │   ├── scene_detection_action.py
│       │   ├── utils.py
│       │   └── visualization.py
│       └── point_cloud_utils_ros.cpp
└── setup.py