Yolo Fighter

Concept

   YOLO Fighter is a 2-player local fighting game powered by real-time computer vision. Instead of pressing buttons. "you become the controller". Strike a pose like your favorite chosen martial arts legend and watch the battle unfold in real-time. Powered by YOLO pose detection, the game reads your body movements through your webcam and translates them into devastating attacks, powerful special moves, or mana-gathering stances. Who will be crowned King of the Match?

Mechanics

• Turn-based combats: players alternate turns, each choosing and performing one of three randomly offered actions (normal attack, power attacks, gather mana). 
• Normal attack: mimic the pose, engage in accuracy gauge mini-game, then stop a moving accuracy gauge with your wrist to deal 0–10 damage.
• Power attack: costs full mana (5). If you have it, it will unleash a guaranteed 20 damage hit.
• Gather mana: hold the mana pose to collect 1–2 mana toward your power attack charge.
• Health: each player starts at 30 HP. The first to reach 0 loses.
• Pose Hold System: you must hold the correct pose for 20 frames to confirm it, to preventing accidental triggers.

How to play

1. Select your camera, and fighter character

2. Follow the instruction on screen. And use your pose to match with one of the action that you're going to use against your opponents. 

3. After your turn end, switch with your friend.

4. The game decide when one of the opponent lose all the health.

Control

Camera selection screen

Action	Key
Scroll up/down through cameras	W / S
Confirm camera selection	Spacebar

Title Screen

Action	Key
Start game	Spacebar
Quit	Q

Character selection screen

Action	Player 1	Player 2
Scroll left	A	4 (numpad)
Scroll right	D	6 (numpad)
Confirm character	Z	Enter
Quit	Q	Q

During Gameplay (Detect Pose / Gauge / Resolve)

Action	Key
Perform action	Mimic pose with your body (held 20 frames)
Stop gauge (Normal Attack)	Move wrist into green STOP circle
End turn	Spacebar or move wrist into green “NEXT TURN” circle
Quit	Q
Toggle Visible Dark Mode	V

Game Over Screen

Action	Key
Restart	R or move wrist into green "RESTART" circle
Quit	Q

Requirements

• OS: Windows
• Python: 3.12+
• Libraries: OpenCV, Ultralytics YOLO, Pygame, NumPy
• Space: 500 MB
• Webcam: Required built-in or standard USB
• Players: 1-2 (local shared screen co-op

Content Advisory

• Simulated martial arts combat (No visible blood or gore)
• Realistic photo-based character references: characters are represented using real martial artist/celebrity photos as pose guides
• Non-commercial educational disclaimer
• Real-time body tracking via webcam (no data is stored or transmitted)
• Suitable for all ages

Privacy Notices

    Screenshots and gameplay footages featured on this page contain images of the development team, included solely to demonstrate the pose detection system. These images are provided for viewing purposes only.    

   Do not reproduce, redistribute, crop, edit or use these images in any form without explicit written permission from the creators.

   This project was created for academic purposes (ITGT523 Computer Vision). All gameplay footage belongs to the development team.   

About us

   This game made by a group of 3 University students as  a part of computer vision project to study the pose detection.  Our aim is to create an innovative turn-based fighting game that make players able to enjoy together and feels challenge.

Question

1. How can I adding more character or pose?
= Please look into "How to add more character and pose tutorial.pdf" included with the zip file.

2. Any plan to update in the future?

=  Maybe, if we have time, we will try make the detection of pose better as well as the tutorial for beginner.

3. Where can I ask question or report the bug that I found?

= We also would like to hear your suggestion. Please feel free to fill out this survey form : https://forms.gle/hN56hv1Yz4EdU17F9