kvnptl / HEART_MET_ODFBM

Object Detection Benchmark

HEART-MET METRICS project | Object Detection Functional Benchmarking

Official repo link: https://github.com/HEART-MET

System config:

• Ubuntu 20.04 LTS
• ROS Noetic
• Python 3

Clone dependent repositories/ros packages (put it in src directory):

1. git clone https://github.com/kvnptl/metrics_refbox.git
2. git clone https://github.com/kvnptl/metrics_refbox_client.git
3. git clone https://github.com/kvnptl/rosbag_recorder.git
4. git clone https://github.com/HEART-MET/metrics_refbox_msgs.git
5. sudo apt install ros-noetic-rospy-message-converter
6. pip3 install sympy

How to run the code:

1. Run object detection code (and metrics refbox client node):
   roslaunch object_detection object_detection_benchmark.launch

2. Publish images from rosbag file (-l means publish continuosly):
   rosbag play -l /home/kvnptl/work/heart_met_competition/bagfiles-001/bagfiles/b-it-bots_2020_11_24_10-17-01.bag

3. Launch Refbox node:
   roslaunch metrics_refbox metrics_refbox.launch

4. Select target object from referee box GUI(checkmark configuration button) and press "Start" to send command to ros node

Debug:

View the rosbag image by running: rosrun image_view image_view image:=/hsrb/head_rgbd_sensor/rgb/image_raw

Data collection

• Distance between table edge and robot central axis is ~90cm.
• Camera position - ~142cm
• Table height ~112cm

Train Yolo model command:

python train.py --img 640 --cfg yolov5s.yaml --hyp hyp.scratch-low.yaml --batch 32 --epochs 150 --data heartmet.yaml --weights yolov5s.pt --workers 24 --name yolo_HeartMet_day2