ace-trump-tech/DeltaForce-OBS-Locker

What it solves

This project provides a real-time object detection system for the game Delta Force, designed to identify and lock onto game characters. It specifically addresses the challenge of distinguishing between actual human players and non-human environmental objects (such as protective suits in the AZ3 Nuclear Power Plant map) to reduce false positives in target acquisition.

How it works

The system uses a visual recognition pipeline based on the YOLOv14 object detection framework. It employs "Game2Real domain adaptation" to align game-rendered features with real-world photography features, allowing the model to reliably identify game characters as "humans."

For the PC version, it captures game frames via OBS and uses Python, OpenCV, and SendInput for simulated mouse movement. The Android version is provided as an APK. The latest V4 update introduces a specific contour filtering layer to isolate and ignore "container protective suits" that were previously misidentified as targets.

Who it’s for

• Python beginners and computer vision enthusiasts interested in learning about real-time object detection.
• Android testers and technical learners studying how visual-based target locking mechanisms work.

Highlights

• YOLOv14 Integration: Uses a specialized real-time detection framework designed for non-ideal imaging conditions.
• Domain Adaptation: Implements Game2Real technology to bridge the gap between game graphics and real-world human detection.
• False Positive Reduction: V4 specifically optimizes for the S10 season's AZ3 map to prevent misidentification of environmental assets.
• C-side implementation: Does not modify game memory, relying solely on image recognition and simulated input APIs.