Topspin Trackers

Repository for The OpenCV International AI Competition 2021 Phase II

Link to the demo

Region

Central Asia + Southern Asia (Bangalore, India)

Prizes

• Regional winners
• Popular Vote winners

Team Members

• Kaustubh Kulkarni
• Rohan Jamadagni
• Jeffrey Paul
• Sucheth Shenoy

---

System Setup

3 OAK-Ds are required for the setup.

First,

git clone https://github.com/Rohansjamadagni/topspin-tracker.git
cd topspin-tracker

Time Synchronization Setup

The times must be synced for all the client devices through NTP for accurate splitting using vibration sensors and stroke classification.

First install ntp on all the devices then enable and start the service.

sudo apt install ntp
sudo systemctl enable --now ntp

Choose any one of the clients as the host, navigate to /etc/ntp.conf , Remove any servers if present and add

server 127.127.1.0
fudge 127.127.1.0 stratum 10

Restart the ntp service

sudo systemctl restart ntp

On the other devices add the ip address of the host previously selected as a server to /etc/ntp.conf

server ip.of.host.server

The time should be synchronised on startup, To sync the Time instantly, run the following commands on all the client devices,

sudo systemctl stop ntp
sudo ntpd -gq
sudo systemctl start ntp

Now we can verify if the time has been synchronized by checking the delay using this command on all the client devices

ntpq -pn

If the offset is less than 0.5ms we are good to go!

OAK-D 1 and OAK-D 2

These OAK-Ds must be mounted on the net clamps in such a way that the cameras face the player being analysed.

On the systems interfacing with the OAK-Ds, run

pip install -r pose_estimation/requirements.txt

Connect 2 vibration sensors to one of the systems and place the sensors under the table, one on each side.

OAK-D 3

This OAK-D must be placed at an elevated umpire's view.

On the system interfacing with this OAK-D, run

pip install -r ball_detection/requirements.txt

Server

On the server, run

sudo apt install ansible
pip install -r server/requirements.txt

Run System

To run the application, first the 3 systems should be connected to the same network.

Open the hosts.yml file in the ansible folder. The IP addresses for the 3 systems must be entered here. Make sure that the IP address of the system entered for pose_estimator_1 also has the vibration sensors connected to it.

Next, the ips.json file must be updated with the host machine's , i.e. the server's IP address in the broker_ip field.

Once this is done, run

ansible-playbook -i ansible/hosts.yml ansible/init.yml
python3 server/subscriber.py

This will initialise all systems with the required code and folder structure.

You're all done! The following command will start the graphical application to record and analyse your sessions.

python3 server/app.py