Package Summary
An object Detection Example for Leo Rover.
- Maintainer status: maintained
- Maintainer: Fictionlab <support AT fictionlab DOT pl>
- Author: Aleksander Szymański <aleks AT fictionlab DOT pl>
- License: MIT
- Source: git https://github.com/LeoRover/leo_examples.git (branch: master)
Note: You can also check a detailed integration tutorial on LeoRover Tech - Object detection
Contents
Overview
This package contains a ROS node which allows to perform object detection on Leo Rover. There is also the default configuration included which contains:
- File with labels
- Neural network models
- Config file with defined colors for specified labels
It is one of a couple of packages made for showing an example usage of a stock Leo Rover.
Usage
Running object detection with the default configuration
If you want to run object detection with the default configuration, simply type:
roslaunch leo_example_object_detection detector.launch
Running object detection with custom configuration
The provided launch file has 4 arguments, which let you choose a neural network model, files with color definitions and label names, as well as camera view. The arguments are as follows:
- labels
- The path to the file with the labels (one label in one line), for the neural network model.
IMPORTANT: If the labels in the file won't be suited for the model, the program might not work properly.
- camera
- The name of the topic with the camera image - you can specify which camera view you want to use for detecting objects.
- model
- The path to the neural network model. There are already two models given (both need the COCO labels), but you can provide a model of your choice.
- config_file
- The path to the config file with the ROS params defining colors for specified labels.
Note: See the ROS API section for more details about the parameters.
Therefore, an example usage with custom configuration looks like this:
roslaunch leo_example_object_detection detector.launch config_file:=<path_to_config> model:=<path_to_model> labels:=<path_to_labels> camera:=<topic_name>
ROS API
object_detector
Subscribed Topics
camera/image_raw (sensor_msgs/Image)- Camera images used for object detection.
Published Topics
detections/compressed (sensor_msgs/CompressedImage)- An image with labels and boxes drawn around the detected objects (an object is detected only when the model's confidence is above the threshold specified by the parameter).
Parameters
Parameters specified in a yaml file. An example values for the parameters are available in the label_colors.yaml file in the config directory.
- The labels for which the color is specified in the colors parameter. The first label will have the first color, the second label the second color etc.
- The colors in RGB format - each inner list is a color definition. The first color will be assigned to the first label from the labels parameter, the second color to the second label etc.
Parameters controlled by the dynamic_reconfigure package:
- The confidence threshold for the model, for showing the detected objects.
