Cubicle Detection User Guide

1. Configuration

Configuration file lies in config/ directory, stored in yaml format.

  1. In cv.yaml

    • ROS topic names for output ROS messages from cubicle_detect

    • image_view: includes parameters regard to showing the results in images and console output

  2. In params.yml

    • parameters required for Superpixel segmentation required for Sailiency based negative obstacle detection.

    • superpixelNum: number of superpixels to generate

  3. In yolov3-spp.yaml

    • Contains information regarding to yolo natwork.

    • config_file: cfg filename

    • weight_file: weight filename

    • threshold: threshold to disregard detection (detections having less detection confidence than this value, are disregarded)

    • detection_classes: detection class label list for the trained network

    • compact_classes: detection class label list we defined

  • Folder yolo_network_config/ contains cfg and weight files.
    • In cfg/yolov3-spp.cfg

      • Network architecture configuration file for Yolov3.

      • You can alter width=416 and height=416 parameters. But the value should be dividable by 32. Also try to maintain the aspect ratio.

    • weights/yolov3-spp.weights with the same name with cfg file is the weight file we have been using.

2. Weight File

Check if the weight file yolov3-spp.weights is already in darknet_ros/yolo_network_config/weights/.

If not, download the weight file and store it in yolo_network_config/weights/

3. Launch File

ROS Launch file lies in launch/.

Excute the command below to only run cubicle_detection; (you might need to run a roasbag or access to a live sensor data feed)

roslaunch cubicle_detect bus.launch

Excute the command below to run cubicle_detection, sslam and cubicle_merge; (you might need to run a roasbag or access to a live sensor data feed)

roslaunch cubicle_detect demo.launch