一、效果展示
此次只选录了以下五种手势,当然你可以自己选择增加手势。
二、项目实现原理
首先通过opencv的手部检测器检测出我们的手,然后录入自己想要检测的手部信息,使用Tensorflow训练得到预训练权重文件(此处已经训练完成,直接调用即可!),调用预训练权重文件对opencv检测的手部信息进行预测,实时返回到摄像头画面,到此整体项目已经实现,此外还可以添加语音模块如speech,对检测到的手势信息进行语音播报。
三、项目环境安装
首先python的版本此处选择为3.7.7(其余版本相差不大的都可)
然后,我们所需要下载的环境如下所示,你可以将其存为txt格式直接在终端输入(具体格式如下图):
pip install -r environment.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
absl-py==1.2.0
attrs==22.1.0
cvzone==1.5.6
cycler==0.11.0
fonttools==4.37.4
kiwisolver==1.4.4
matplotlib==3.5.3
mediapipe==0.8.9.1
numpy==1.21.6
opencv-contrib-python==4.6.0.66
opencv-python==4.6.0.66
opencv-python-headless==4.6.0.66
packaging==21.3
Pillow==9.2.0
protobuf==3.19.1
pyparsing==3.0.9
python-dateutil==2.8.2
six==1.16.0
speech==0.5.2
typing_extensions==4.4.0
保存格式如下:
四、代码实现
模型预训练权重如下
点击这里下载
import cv2
from cvzone.HandTrackingModule import HandDetector
from cvzone.ClassificationModule import Classifier
from PIL import Image, ImageDraw, ImageFont
import numpy as np
import math
import time
# import speech
cap = cv2.VideoCapture(0)
cap.set(3, 1280)
cap.set(4, 720)
detector = HandDetector(maxHands=1)
classifile = Classifier("./model/keras_model.h5", "./model/labels.txt")
offset = 20
imgSize = 300
counter = 0
labels = ['666', '鄙视', 'Good', '比心', '击掌', '握拳']
# folder = r"F:\opencv_game\HandSignDetection\Data\Love"
while True:
success, img = cap.read()
img = cv2.flip(img, 1)
imgOutput = img.copy()
hands, img = detector.findHands(img)
if hands:
hand = hands[0]
x, y, w, h = hand['bbox']
imgWhite = np.ones((imgSize, imgSize, 3), np.uint8)*255
imgCrop = img[y - offset:y + h + offset, x - offset:x + w + offset]
imgCropShape = imgCrop.shape
aspectRatio = h/w
if aspectRatio > 1:
k = imgSize/h
wCal = math.ceil(k*w)
imgResize = cv2.resize(imgCrop, (wCal, imgSize))
imgResizeShape = imgResize.shape
wGap = math.ceil((imgSize - wCal)/2)
imgWhite[:, wGap:wCal+wGap] = imgResize
prediction, index = classifile.getPrediction(imgWhite)
print(prediction, index)
else:
k = imgSize / w
hCal = math.ceil(k * h)
imgResize = cv2.resize(imgCrop, (imgSize, hCal))
imgResizeShape = imgResize.shape
hGap = math.ceil((imgSize - hCal) / 2)
imgWhite[hGap:hCal + hGap,:] = imgResize
prediction, index = classifile.getPrediction(imgWhite)
# 解决cv2.putText绘制中文乱码
def cv2AddChineseText(img, text, position, textColor=(255, 255, 255), textSize=50):
if (isinstance(img, np.ndarray)): # 判断是否OpenCV图片类型
img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
# 创建一个可以在给定图像上绘图的对象
draw = ImageDraw.Draw(img)
# 字体的格式
fontStyle = ImageFont.truetype(
"simsun.ttc", textSize, encoding="utf-8")
# 绘制文本
draw.text(position, text, textColor, font=fontStyle)
# 转换回OpenCV格式
return cv2.cvtColor(np.asarray(img), cv2.COLOR_RGB2BGR)
cv2.rectangle(imgOutput, (x - offset, y - offset - 50),
(x-offset+130, y-offset), (255, 0, 255), cv2.FILLED)
# cv2.putText(imgOutput, labels[index], (x,y-24),
# cv2.FONT_HERSHEY_COMPLEX, 1.5, (255, 255, 255), 2)
# 中文
img = cv2AddChineseText(imgOutput, labels[index], (x - offset, y - offset - 50))
cv2.rectangle(img, (x-offset, y-offset),
(x+w+offset, y+h+offset), (255,0,255),4)
# speech.say(labels[index])
# cv2.imshow('ImageCrop', imgCrop)
# cv2.imshow('ImageWhite', imgWhite)
cv2.imshow('Image', img)
key = cv2.waitKey(1)
if key == ord('s'):
pass
elif key == 27:
break
五、总结
到此这篇关于利用OpenCV+Tensorflow实现手势识别的文章就介绍到这了,更多相关OpenCV+Tensorflow手势识别内容请搜索编程网以前的文章或继续浏览下面的相关文章希望大家以后多多支持编程网!