Attempts to mark/identify objects(players) both enemies and teammates using YoloV8's object detection on a custom dataset.
- Please consult Requirements first.
- Read all TODO comments in Program.cs (Replace ModelPath to a absolute path!)
- Attempt to run (Running in a debug state will decrease performance exponentially!)
- If running it'll attempt to take a full screenshot of your desktop (Monitor 0) every couple of milliseconds & process it through yolov8.
The model provided in onnx format is a YoloV8 small model trained on 1.25k images at 1920x1088 for about 425~ epochs (Original's are 1920x1080).
Model dataset will NOT be released!
- Windows Only
- Nvidia GPU
- Powerful CPU (Anything within the last couple of years)
- CUDA 11.X, cuDNN 11.X, zlib 1.2.3 installed / linked to path
- Knowledge
- YoloV8.NET
- Emgu.CV
- Emgu.CV.Bitmap
- Emgu.CV.Runtime.Windows
- Microsoft.ML.OnnxRuntime.GPU
- System.Drawing.common
This proof of concept is not viable for cheating since it has very bad predictions (camouflaged players, and high prediction delay 50-80ms). please do not cheat on battlebit :)