Introduction

This is a YOLOV7 based APEX Aimbot

Features

• Enemy and Friend Differentiation: The model is equipped with the capability to distinguish between enemies and friends, enabling real-time responsive actions. Whether it's on a military battlefield or in security surveillance, it can automatically identify potential threats and harmless targets, thereby greatly enhancing decision-making speed and accuracy.

• PID Smooth Moving: Utilizing a Proportional-Integral-Derivative (PID) algorithm, the model ensures smooth and stable movement trajectories. This feature is especially useful in target tracking or precision operations, effectively eliminating jitters caused by operational delays or hardware constraints.

• Real-Time Detection Results Display: The system displays detection results in real-time through an independent process, allowing users to obtain key information instantly. This not only improves the user experience but also provides timely data support in emergency situations.

• Personalized Settings Through Config File: Users can personalize the model settings by editing the config file, such as detection sensitivity, alert thresholds, etc., achieving applications that better align with individual or organizational needs.

• TensorRT Speed Up: With the application of TensorRT technology, the model significantly boosts its running speed and effectively solves the shaking problem, particularly when operating at high speeds.

• Model Encryption: The system offers encryption services for ONNX and TRT models, ensuring the safety and uniqueness of the model and preventing potential theft and tampering.

• Screenshot Saving During Locking or Detection: The system automatically saves screenshots when a target is locked or an anomaly is detected. This can be used for subsequent analysis and also for collecting a new dataset that includes both false positives and negatives.

• Image Annotation Using Current Models: By annotating images using the current models, the speed of data annotation is significantly increased, further boosting the efficiency of model training.

1.2. Environment set up in Windows10 & Windows11 版本对齐

The following method has being tested and successed under Windows 10 Pro Version 21H2/22H2, Windows11 Pro Version 22H2 and Windows11 Pro Insider View Build 25346. Technically, it works under all latest Windows OS builds.

• Version-align

  CUDA	cuDNN	TensorRT	PyTorch
  11.7.0	8.5.0	8.5.2.2	2.0.0
  11.8.0	8.6.0	8.5.3.1	2.0.0
  ...	...	...	...

  We will be using the second row as our package manifests.

Environment

My envrionment uses python3.7.8 #3.7.8

conda create -n apex python=3.7            #先下载anaconda，然后添加环境变量到电脑中
conda activate apex       # (或者是C:\ProgramData\anaconda3\Scripts\activate.bat apex)
pip install pipwin                         
############
（set http_proxy=http://127.0.0.1:10809     #加入代理 填写自己的代理设置
set https_proxy=http://127.0.0.1:10809
pipwin install pycuda）
##################
pip install -r requirements.txt        #程序自动会一个个的安装，如果出错需要先解决错误再进行后续的安装

1.安装CUDA。

（也可以按照官方说明操作CUDA official website：）。Install cuda11.8 with tensorrt following the NVIDIA official instructions

2.安装cuDNN:

• 注册NVIDIA developer program. NVIDIA developer program.
• 转到 cuDNN 下载站点：cuDNN download archive。cuDNN download archive.
• 单击Download cuDNN v8.6.0 (August 8th, 2022), for CUDA 11.x。
• 下载Local Installer for Windows (Zip).
• 解压缩cudnn-windows-x86_64-8.5.0.96_cuda11-archive.zip。
• 复制所有三个文件夹 ( bin,, ) 并将它们粘贴到include安装目录中。（注意、、文件夹已存在于 CUDA 文件夹中。）。libCUDAC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.7binincludelib #############################

3.安装PyTorch。

conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia

4.安装TensorRT。

• 转到TensorRT 下载站点。 TensorRT download site.
• 下载TensorRT 8.5 GA for Windows 10 and CUDA 11.0, 11.1, 11.2, 11.3, 11.4, 11.5, 11.6, 11.7 and 11.8 ZIP Package.
• TensorRT-8.5.2.2从 中解压文件夹TensorRT-8.5.2.2.Windows10.x86_64.cuda-11.8.cudnn8.6.zip。
• 添加\TensorRT-8.5.2.2\lib到 PATH 环境变量中。
• 转到文件夹\TensorRT-8.5.2.2\python
• 在命令窗口中输入

conda activate yolov8      # activate dedicated environment

pip install tensorrt-8.5.2.2-cp310-none-win_amd64.whl    # install tensorrt package to python

5.安装 python 要求。

pip install -r requirements.txt   （若出现错误则先解决错误，再重复运行此代码）

6.#要点：

6.1zlibwapi.dll缺失

将C:\Program Files\NVIDIA Corporation\Nsight Systems 2022.4.2\host-windows-x64\zlib.dll 我将其复制并重命名为：C:\Program Files\NVIDIA GPUComputing Toolkit\CUDA\v11.8\bin\zlibwapi.dll

7.#Run

Running for apex (default hold left/right button to auto aim, side button(x2) to auto aim and shoot, side button(x1) to enable and disable the AI:

python apex.py

8.验证环境插件有效性

Verify installation and check versions.

• Verify installation of CUDA, cuDNN, PyTorch and TensorRT.

  • Verify CUDA.

    nvcc -V

    If installment successed, you should see prompts like:

    nvcc: NVIDIA (R) Cuda compiler driver
    Copyright (c) 2005-2022 NVIDIA Corporation
    Built on Tue_May__3_19:00:59_Pacific_Daylight_Time_2022
    Cuda compilation tools, release 11.7, V11.7.64
    Build cuda_11.7.r11.7/compiler.31294372_0

  • Verify cuDNN.

    python
    import torch
    print(torch.backends.cudnn.version())

  • Verify PyTorch.

    python
    import torch
    print(torch.__version__)

  • Verify TensorRT.

    pip show tensorrt

    If installment successed, you should see prompts like:

    Name: tensorrt
    Version: 8.5.2.2
    Summary: A high performance deep learning inference library
    Home-page: https://developer.nvidia.com/tensorrt
    Author: NVIDIA Corporation

You can get the customized settings in configs/apex.yaml, set your suitable smooth hyperparameter

Annotate the dataset using current model

python utils/anno_imgs.py --data_dir your_dataset_dir --engine_path your_trt_engine_path