curly

A proof-of-concept system that leverages the YOLO object detection algorithm to identify characters in real time.

Screen Detection

This project is a proof-of-concept application that uses the YOLO (You Only Look Once) object detection algorithm to identify and track human targets directly from the screen in real time. It captures a region around the cursor, processes the image using a lightweight YOLOv4-tiny model via OpenCV’s DNN module, and returns the most confident detection.

Requirements

System

• Windows 10/11 (64-bit)
• ~200–300 MB free disk space (for model weights and dependencies)

Development Environment

• C++17 compatible compiler (e.g., Visual Studio 2019 or newer)
• Windows SDK

Libraries & Dependencies

• OpenCV (version 4.x recommended, with DNN module enabled)
• DirectX 11 (included with Windows SDK)
• Dear ImGui (for UI)

Required Libraries (Linker)

• d3d11.lib
• user32.lib
• gdi32.lib

Use Cases

• Triggerbots externally (without touching memory)
• Aimbots/Aimlocks

Excuse me for the code, it's "bad".