Vision System

Installation

Clone the repository

git clone https://github.com/Nicolasalan/Perception-System.git

Usage

To start detection

cd <catkin_workspace>/
source devel/setup.bash

start detection

Before starting the vision system, it is necessary to modify the main_config.yaml file so that the system can find the configuration files. To do this, just modify the path to the vision system configuration file. The vision system configuration file is located at:

<catkin_workspace>/src/Perception-System/dodo_detector_ros/config/main_config.yaml

The main parameters that must be modified are:

# ['sift', 'rootsift', 'tf1', 'tf2']
detector_type: tf2

saved_model: ~/dodo_detector_ros/saved_model

# path of file saved_model.pb
inference_graph: ~/dodo_detector_ros/saved_model/saved_model.pb

# file path label_map.pbtxt
label_map: ~/dodo_detector_ros/saved_model/label_map.pbtxt

# minimum confidence for detection
tf_confidence: 0.9 

sift_min_pts: 10

sift_database_path: ~/.dodo_detector_ros/sift_database

# local frame where detection will be performed
global_frame: kinect_one_depth

After modifying the main_config.yaml file, just start the system view with the command:

roslaunch dodo_detector_ros detect_kinect.launch

Follow this path in Rviz to see the detection in real time

|-- rviz
    |-- add
        |-- by topic
            |-- dodo_detector_ros
                |-- labeled_image
                    |-- image

Initialize object location

To initialize the object location system, you first have to modify the objects.py file so that the coordinates have a local reference. To do this, just modify the objects.py file, which is located at:

<catkin_workspace>/src/Perception-System/hera_objects/src/objects.py

Inside the file there is a parameter called reference_frame, which only places the frame in which the object will have as a reference. After modifying the file, just start the object location system with the command:

roslaunch hera_objects objects.launch

Call the service to see the positions

rosservice call /objects "condition: 'closest'"

Example of output:

position: 
  x: 0.0
  y: 0.0
  z: 0.0
  rx: 0.0
  ry: 0.0
  rz: 0.0

Links to related tutorials

• freenect_stack
• dodo_detector_ros