RN_AI

🧠 Overview

RN_AI is a high-performance YOLO-based AI assistant with an intuitive DearPyGui interface. Built with Kalman filtering, and optimized for competitive gaming.

• Entry point: main.py
• Core logic: src/ and core/
• Inspired by: RN_AI_cpp

💡 The RN_AI_cpp project is solid and well-designed, but this Python pipeline delivers even higher FPS than C++.

📊 Performance Comparison (RTX 2060)

All benchmarks performed on NVIDIA RTX 2060 - screen capture excluded.

📈 RN_AI Profiling Results

┌─────────────────────────────┐
│ [PROFILE] RTX 2060          │
├─────────────────────────────┤
│ Capture  : 0.09 ms          │
│ Preproc  : 0.00 ms          │
│ Infer    : 3.12 ms          │
│ Post     : 0.16 ms          │
├─────────────────────────────┤
│ Total    : 3.37 ms ✓        │
└─────────────────────────────┘

🎯 Key Benefits

While RN_AI has slightly higher end-to-end latency, it delivers:

• ✅ Significantly higher FPS (~240 vs ~140)
• ✅ Smoother tracking and motion
• ✅ Higher temporal resolution
• ✅ Flexible experimentation in Python
• ✅ Rapid iteration capability

⚠️ Important Warnings

⚡ Quick Start

Installation & Setup

install.bat

python src/installer.py

run.bat

🖥️ GUI Preview

Intuitive interface with real-time monitoring and configuration:

System Panel	Aim System	Class Detection

Driver Control	Bypass Settings	Strafe Config	Config Manager

📁 Project Structure

RN_AI/
├── main.py                  # Entry point - Start here
├── cfg.json                 # Main configuration file
├── requirements.txt         # Python dependencies
├── *.bat                    # Helper scripts (build, install, run)
│
├── src/                     # Inference & Utilities
│   ├── app.py              # Application core
│   ├── inference_engine.py # YOLO inference pipeline
│   ├── screenshot_manager.py # Screen capture
│   ├── pid.py              # PID controller
│   ├── profiler.py         # Performance profiling
│   └── ...
│
├── core/                    # GUI & Control Logic
│   ├── gui.py              # DearPyGui interface
│   ├── aiming.py           # Aiming algorithms
│   ├── recoil.py           # Recoil compensation
│   ├── config.py           # Config management
│   ├── input.py            # Input handling
│   └── ...
│
└── docs/                    # Documentation & Assets
    ├── demo.gif            # Demo animation
    ├── gui/                # GUI screenshots
    └── HELP.md             # Help documentation

🧪 Profiling & Performance Analysis

CPU Profiling

Analyze CPU usage patterns with py-spy:

profile.bat

This generates detailed CPU profiling data to identify bottlenecks and optimization opportunities.

🤖 Model Support

RN_AI supports multiple YOLO model formats and versions:

Supported YOLO Versions

• ✅ YOLO v8 through v12
• 📁 Example models available in .weight/
• 🎛️ Fully configurable from GUI

📝 License & Credits

• Inspired by: RN_AI_cpp
• Framework: DearPyGui
• AI Model: YOLO Object Detection

✨ Features

• Makcu support with optional input masking
• Kalman filtering & prediction
• Dead-zone logic (no target jumping)
• Multiple capture sources:
  • Screen
  • OBS
  • Capture Card
• Class filtering:
  • Enable / disable
  • Custom class sets
  • Per-class aim positions

🎥 Capture

• Sources: Standard (screen), OBS, Capture Card.
• Capture offsets can be set in the GUI.

🧷 Config & Classes

• Stored in cfg.json. Auto-generated if missing.
• Class names can be loaded or entered manually in the GUI.
• Per-class confidence/IOU is supported.

Documentation 📘

Full GUI and parameter help