关于yolov5进行FPS游戏的目标检测，实现自动瞄准。_python+yolov5 fps训练

前言

YOLOV5是一个基于视觉识别的开源项目，本人制作的目的是研究YOLO的应用，并非制作游戏外挂。

类似效果

没时间进行游戏实测画面，暂时放一个其他人的效果吧
【APEX目标识别 2000张图的训练结果-哔哩哔哩】

main.py代码

import argparse
import os
import time

import cv2
import numpy as np
import pynput
import torch
import win32con
import win32gui
from simple_pid import PID

from aim_csgo.apex_aim import lock
from aim_csgo.cs_model import load_model
from aim_csgo.screen_inf import (get_parameters, grab_screen_mss,
                                 grab_screen_win32)
from aim_csgo.verify_args import verify_args
from utils.augmentations import letterbox
from utils.general import non_max_suppression, scale_coords, xyxy2xywh

parser = argparse.ArgumentParser()
parser.add_argument('--model-path', type=str, default='weights/best.pt', help='模型位置 model address')
parser.add_argument('--imgsz', type=int, default=640, help='和训练模型时imgsz一样')
parser.add_argument('--conf-thres', type=float, default=0.1, help='置信度阈值')
parser.add_argument('--iou-thres', type=float, default=0.45, help='交並比閥值')
parser.add_argument('--use-cuda', type=bool, default=False, help='是否使用cuda') 
parser.add_argument('--show-window', type=bool, default=True, help='是否显示检测窗口(debug用,若是True,不要去點右上角的X)')
parser.add_argument('--top-most', type=bool, default=True, help='是否保持检测窗口在最上层')
parser.add_argument('--resize-window', type=float, default=1/2, help='缩放窗口大小')
parser.add_argument('--thickness', type=int, default=2, help='边框线条粗细，需大于1/resize-window')
parser.add_argument('--show-fps', type=bool, default=False, help='是否显示fps')
parser.add_argument('--show-label', type=bool, default=True, help='是否显示标签')
parser.add_argument('--use_mss', type=str, default=True, help='是否使用mss截屏；为False時使用win32截屏；笔记本用True')
parser.add_argument('--region', type=tuple, default=(0.4, 0.4), help='检测范围；分别为x轴和y轴，(1.0, 1.0)表示全屏检测，越低检测值范围越小(以屏幕中心为检测中心)')
parser.add_argument('--hold-lock', type=bool, default=False, help='lock模式；True为按住，False为切換')
parser.add_argument('--lock-sen', type=float, default= 3.0, help='lock幅度系數,遊戲中灵敏度(建议不要调整)')
parser.add_argument('--lock-smooth', type=float, default=1.9, help='lock平滑系数；越大越平滑')
parser.add_argument('--lock-button', type=str, default='right', help='lock按键；只支持鼠标按键，调整自瞄键[left, middle, right, x1, x2]')
parser.add_argument('--head-first', type=bool, default=True, help='是否优先检测头部')
parser.add_argument('--lock-tag', type=list, default=[0], help='对应标签；person(倘若模型不同，请自行修改对应标签)')
parser.add_argument('--lock-choice', type=list, default=[0], help='目标选择；决定锁定的目标，从自己标签选择')
args = parser.parse_args()
'------------------------------------------------------------------------------------'
verify_args(args)
cur_dir = os.path.dirname(os.path.abspath(__file__)) + '\\'
args.model_path = cur_dir + args.model_path
args.lock_tag = [str(i) for i in args.lock_tag]
args.lock_choice = [str(i) for i in args.lock_choice]
device = 'cuda' if args.use_cuda else 'cpu'
half = device != 'cpu'
imgsz = args.imgsz
conf_thres = args.conf_thres
iou_thres = args.iou_thres
top_x, top_y, x, y = get_parameters()
len_x, len_y = int(x * args.region[0]), int(y * args.region[1])
top_x, top_y = int(top_x + x // 2 * (1. - args.region[0])), int(top_y + y // 2 * (1. - args.region[1]))
monitor = {'left': top_x, 'top': top_y, 'width': len_x, 'height': len_y}
model = load_model(args)
stride = int(model.stride.max())
names = model.module.names if hasattr(model, 'module') else model.names
lock_mode = False
team_mode = True
lock_button = eval('pynput.mouse.Button.' + args.lock_button)
mouse = pynput.mouse.Controller()

#pid系数可自行调整(一般960显卡参数：kp = 0.5，ki = 0.17，kd = 0.03)
pidx = PID(0.5, 0.17, 0.03, setpoint=0, sample_time=0.001,)
pidy = PID(1.22, 0.12, 0.0, setpoint=0, sample_time=0.001,)
pidx.output_limits = (-4000 ,4000)
pidy.output_limits = (-3000 ,3000)

if args.show_window:
    cv2.namedWindow('aim', cv2.WINDOW_NORMAL)
    cv2.resizeWindow('aim', int(len_x * args.resize_window), int(len_y * args.resize_window))
def on_click(x, y, button, pressed):
    global lock_mode
    if button == lock_button:
        if args.hold_lock:
            if pressed:
                lock_mode = True
                print('locking...')
            else:
                lock_mode = False
                print('lock mode off')
        else:
            if pressed:
                lock_mode = not lock_mode
                print('lock mode', 'on' if lock_mode else 'off')

listener = pynput.mouse.Listener(on_click=on_click)
listener.start()

print('enjoy yourself!')
t0 = time.time()
cnt = 0

while True:

    if cnt % 20 == 0:
        top_x, top_y, x, y = get_parameters()
        len_x, len_y = int(x * args.region[0]), int(y * args.region[1])
        top_x, top_y = int(top_x + x // 2 * (1. - args.region[0])), int(top_y + y // 2 * (1. - args.region[1]))
        monitor = {'left': top_x, 'top': top_y, 'width': len_x, 'height': len_y}
        cnt = 0

    if args.use_mss:
        img0 = grab_screen_mss(monitor)
        img0 = cv2.resize(img0, (len_x, len_y))
    else:
        img0 = grab_screen_win32(region=(top_x, top_y, top_x + len_x, top_y + len_y))
        img0 = cv2.resize(img0, (len_x, len_y))

    img = letterbox(img0, imgsz, stride=stride)[0]

    img = img.transpose((2, 0, 1))[::-1]
    img = np.ascontiguousarray(img)

    img = torch.from_numpy(img).to(device)
    img = img.half() if half else img.float()
    img /= 255.

    if len(img.shape) == 3:
        img = img[None]

    pred = model(img, augment=False, visualize=False)[0]
    pred = non_max_suppression(pred, conf_thres, iou_thres, agnostic=False)

    aims = []
    for i, det in enumerate(pred):
        gn = torch.tensor(img0.shape)[[1, 0, 1, 0]]
        if len(det):
            det[:, :4] = scale_coords(img.shape[2:], det[:, :4], img0.shape).round()

            for *xyxy, conf, cls in reversed(det):
                # bbox:(tag, x_center, y_center, x_width, y_width)
                xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
                line = (cls, *xywh)  # label format
                aim = ('%g ' * len(line)).rstrip() % line
                aim = aim.split(' ')
                aims.append(aim)

        if len(aims):
            if lock_mode:
                lock(aims, mouse, top_x, top_y, len_x, len_y, args, pidx, pidy)

        if args.show_window:
            for i, det in enumerate(aims):
                tag, x_center, y_center, width, height = det
                x_center, width = len_x * float(x_center), len_x * float(width)
                #print("width:" , width)
                #print("x_center:", x_center)
                y_center, height = len_y * float(y_center), len_y * float(height)
                top_left = (int(x_center - width / 2.), int(y_center - height / 2.))
                #print("top_left:", top_left)
                bottom_right = (int(x_center + width / 2.), int(y_center + height / 2.))
                #print("bottom_right:", bottom_right)
                cv2.rectangle(img0, top_left, bottom_right, (0, 255, 0), thickness=args.thickness)
                if args.show_label:
                    cv2.putText(img0, tag, top_left, cv2.FONT_HERSHEY_SIMPLEX, 0.7, (235, 0, 0), 4)

    if args.show_window:
        if args.show_fps:
            cv2.putText(img0,"FPS:{:.1f}".format(1. / (time.time() - t0)), (10, 50), cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 0, 235), 4)
            # cv2.putText(img0, "lock:{:.1f}".format(lock_mode), (10, 100), cv2.FONT_HERSHEY_SIMPLEX, 2,(0, 0, 235), 4)
            # cv2.putText(img0, "team:{:.1f}".format(team_mode), (10, 150), cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 0, 235), 4)
            print(1. / (time.time() - t0))
            t0 = time.time()

        cv2.imshow('aim', img0)

        if args.top_most:
            hwnd = win32gui.FindWindow(None, 'aim')
            CVRECT = cv2.getWindowImageRect('aim')
            win32gui.SetWindowPos(hwnd, win32con.HWND_TOPMOST, 0, 0, 0, 0, win32con.SWP_NOMOVE | win32con.SWP_NOSIZE)

        cv2.waitKey(1)
    pidx(0)
    pidy(0)
    cnt += 1

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180

使用方法

1.下载安装罗技驱动，并不需要买一个罗技鼠标。
2.安装必要的python库
3.运行即可

注意

修改这个代码可以选择是否开启和关闭功能

parser.add_argument('--lock-button', type=str, default='right', help='lock按键；只支持鼠标按键，调整自瞄键[left, middle, right, x1, x2]')
1

完整链接↓

知识付费按钮