Autonavigator for Zombie Rush

This autonavigator automatically set's cursor on position of detected zombies and then you can click to shoot.

Dataset - https://universe.roboflow.com/test-o0ilg/zombie-rush

Autolabeling

You can use autolabel.py to autolabel images at the images folder and then upload them to roboflow, with the labels.txt file copied there.

Keys

P - Pause
R - Resume
F - Save screenshot to images folder for dataset

Installation on Windows using Conda

Follow this command order in conda in order to install everything as I did

1. Create conda environment

conda create -n autonavigator python=3.10
conda activate autonavigator

2. Install cudatoolkit and cudnn

conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0

3. Install pytorch with cuda

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

4. Open project folder

cd /d /path/to/folder

4. Install packages from requirements.txt

pip install -r requirements.txt

5. Open bot folder

cd bot

6. Start bot using raw command or using start.bat if you are on Windows

python.exe .\main.py --device=0 --weights .\weights\yolov8_zombie_m_v3.pt

Windows

start.bat

Enhancements that you can make

1. You can tune DXCam to wait for last frame available, and not spam grab. Or maybe use other libraries
2. Export yolo model using TensorRT to optimize it's execution on GPU
3. Add images to dataset and train yolo model on it to make detection more accurate

Preview

Credits

Andrew Wong for RobloxBot - bot that collects coins in Murder Mystery 2, used as a base, check it out