【CanMV K230 AI视觉】 人体检测

动态测试效果可以去下面网站自己看。

B站视频链接：已做成合集
抖音链接：已做成合集

---

人体检测

人体检测是判断摄像头画面中有无出现人体，常用于人体数量检测，人流量监控以及安防监控等。

'''
实验名称：人体检测
实验平台：01Studio CanMV K230
教程：wiki.01studio.cc
'''from libs.PipeLine import PipeLine, ScopedTiming
from libs.AIBase import AIBase
from libs.AI2D import Ai2d
import os
import ujson
from media.media import *
from time import *
import nncase_runtime as nn
import ulab.numpy as np
import time
import utime
import image
import random
import gc
import sys
import aicube# 自定义人体检测类
class PersonDetectionApp(AIBase):def __init__(self,kmodel_path,model_input_size,labels,anchors,confidence_threshold=0.2,nms_threshold=0.5,nms_option=False,strides=[8,16,32],rgb888p_size=[224,224],display_size=[1920,1080],debug_mode=0):super().__init__(kmodel_path,model_input_size,rgb888p_size,debug_mode)self.kmodel_path=kmodel_path# 模型输入分辨率self.model_input_size=model_input_size# 标签self.labels=labels# 检测anchors设置self.anchors=anchors# 特征图降采样倍数self.strides=strides# 置信度阈值设置self.confidence_threshold=confidence_threshold# nms阈值设置self.nms_threshold=nms_thresholdself.nms_option=nms_option# sensor给到AI的图像分辨率self.rgb888p_size=[ALIGN_UP(rgb888p_size[0],16),rgb888p_size[1]]# 显示分辨率self.display_size=[ALIGN_UP(display_size[0],16),display_size[1]]self.debug_mode=debug_mode# Ai2d实例，用于实现模型预处理self.ai2d=Ai2d(debug_mode)# 设置Ai2d的输入输出格式和类型self.ai2d.set_ai2d_dtype(nn.ai2d_format.NCHW_FMT,nn.ai2d_format.NCHW_FMT,np.uint8, np.uint8)# 配置预处理操作，这里使用了pad和resize，Ai2d支持crop/shift/pad/resize/affine，具体代码请打开/sdcard/app/libs/AI2D.py查看def config_preprocess(self,input_image_size=None):with ScopedTiming("set preprocess config",self.debug_mode > 0):# 初始化ai2d预处理配置，默认为sensor给到AI的尺寸，您可以通过设置input_image_size自行修改输入尺寸ai2d_input_size=input_image_size if input_image_size else self.rgb888p_sizetop,bottom,left,right=self.get_padding_param()self.ai2d.pad([0,0,0,0,top,bottom,left,right], 0, [0,0,0])self.ai2d.resize(nn.interp_method.tf_bilinear, nn.interp_mode.half_pixel)self.ai2d.build([1,3,ai2d_input_size[1],ai2d_input_size[0]],[1,3,self.model_input_size[1],self.model_input_size[0]])# 自定义当前任务的后处理def postprocess(self,results):with ScopedTiming("postprocess",self.debug_mode > 0):# 这里使用了aicube模型的后处理接口anchorbasedet_post_preocessdets = aicube.anchorbasedet_post_process(results[0], results[1], results[2], self.model_input_size, self.rgb888p_size, self.strides, len(self.labels), self.confidence_threshold, self.nms_threshold, self.anchors, self.nms_option)return dets# 绘制结果def draw_result(self,pl,dets):with ScopedTiming("display_draw",self.debug_mode >0):if dets:pl.osd_img.clear()for det_box in dets:x1, y1, x2, y2 = det_box[2],det_box[3],det_box[4],det_box[5]w = float(x2 - x1) * self.display_size[0] // self.rgb888p_size[0]h = float(y2 - y1) * self.display_size[1] // self.rgb888p_size[1]x1 = int(x1 * self.display_size[0] // self.rgb888p_size[0])y1 = int(y1 * self.display_size[1] // self.rgb888p_size[1])x2 = int(x2 * self.display_size[0] // self.rgb888p_size[0])y2 = int(y2 * self.display_size[1] // self.rgb888p_size[1])if (h<(0.1*self.display_size[0])):continueif (w<(0.25*self.display_size[0]) and ((x1<(0.03*self.display_size[0])) or (x2>(0.97*self.display_size[0])))):continueif (w<(0.15*self.display_size[0]) and ((x1<(0.01*self.display_size[0])) or (x2>(0.99*self.display_size[0])))):continuepl.osd_img.draw_rectangle(x1 , y1 , int(w) , int(h), color=(255, 0, 255, 0), thickness = 2)pl.osd_img.draw_string_advanced( x1 , y1-50,32, " " + self.labels[det_box[0]] + " " + str(round(det_box[1],2)), color=(255,0, 255, 0))else:pl.osd_img.clear()# 计算padding参数def get_padding_param(self):dst_w = self.model_input_size[0]dst_h = self.model_input_size[1]input_width = self.rgb888p_size[0]input_high = self.rgb888p_size[1]ratio_w = dst_w / input_widthratio_h = dst_h / input_highif ratio_w < ratio_h:ratio = ratio_welse:ratio = ratio_hnew_w = (int)(ratio * input_width)new_h = (int)(ratio * input_high)dw = (dst_w - new_w) / 2dh = (dst_h - new_h) / 2top = int(round(dh - 0.1))bottom = int(round(dh + 0.1))left = int(round(dw - 0.1))right = int(round(dw - 0.1))return  top, bottom, left, rightif __name__=="__main__":# 显示模式，默认"hdmi",可以选择"hdmi"和"lcd"display_mode="lcd"if display_mode=="hdmi":display_size=[1920,1080]else:display_size=[800,480]# 模型路径kmodel_path="/sdcard/app/tests/kmodel/person_detect_yolov5n.kmodel"# 其它参数设置confidence_threshold = 0.2nms_threshold = 0.6rgb888p_size=[1920,1080]labels = ["person"]anchors = [10, 13, 16, 30, 33, 23, 30, 61, 62, 45, 59, 119, 116, 90, 156, 198, 373, 326]# 初始化PipeLinepl=PipeLine(rgb888p_size=rgb888p_size,display_size=display_size,display_mode=display_mode)pl.create()# 初始化自定义人体检测实例person_det=PersonDetectionApp(kmodel_path,model_input_size=[640,640],labels=labels,anchors=anchors,confidence_threshold=confidence_threshold,nms_threshold=nms_threshold,nms_option=False,strides=[8,16,32],rgb888p_size=rgb888p_size,display_size=display_size,debug_mode=0)person_det.config_preprocess()clock = time.clock()try:while True:os.exitpoint()clock.tick()img=pl.get_frame()  # 获取当前帧数据res=person_det.run(img)  # 推理当前帧person_det.draw_result(pl,res)  # 绘制结果到PipeLine的osd图像print(res) # 打印结果pl.show_image()  # 显示当前的绘制结果gc.collect()print(clock.fps()) #打印帧率#IDE中断注销相关对象，释放资源except Exception as e:sys.print_exception(e)finally:person_det.deinit()pl.destroy()

使用类	说明
PersonDetectionApp	人体检测