终于等到你,还好我没有放弃。最后一部分,sbb,上代码。
增加一个文件labels.txt,内容是我们的标签,注意放置位置。
增加一个inference.ipynb用于编写推理代码,注意放置位置。
from openvino.runtime import Core
import openvino.runtime as ov
import cv2 as cv
import numpy as np
import tensorflow as tf
OpenVINO 模型推理器(class)
class Predictor:
"""
OpenVINO 模型推理器
"""
def __init__(self, model_path):
ie_core = Core()
model = ie_core.read_model(model=model_path)
self.compiled_model = ie_core.compile_model(model=model, device_name="CPU")
def get_inputs_name(self, num):
return self.compiled_model.input(num)
def get_outputs_name(self, num):
return self.compiled_model.output(num)
def predict(self, input_data):
return self.compiled_model([input_data])
def get_request(self):
return self.compiled_model.create_infer_request()
图像预处理
def process_image(input_image, size):
"""输入图片与处理方法,按照PP-Yoloe模型要求预处理图片数据
Args:
input_image (uint8): 输入图片矩阵
size (int): 模型输入大小
Returns:
float32: 返回处理后的图片矩阵数据
"""
max_len = max(input_image.shape)
img = np.zeros([max_len,max_len,3],np.uint8)
img[0:input_image.shape[0],0:input_image.shape[1]] = input_image # 将图片放到正方形背景中
img = cv.cvtColor(img,cv.COLOR_BGR2RGB) # BGR转RGB
img = cv.resize(img, (size, size), cv.INTER_NEAREST) # 缩放图片
img = np.transpose(img,[2, 0, 1]) # 转换格式
img = img / 255.0 # 归一化
img = np.expand_dims(img,0) # 增加维度
return img.astype(np.float32)
图像后处理
def process_result(box_results, conf_results):
"""按照PP-Yolove模型输出要求,处理数据,非极大值抑制,提取预测结果
Args:
box_results (float32): 预测框预测结果
conf_results (float32): 置信度预测结果
Returns:
float: 预测框
float: 分数
int: 类别
"""
conf_results = np.transpose(conf_results,[0, 2, 1]) # 转置
# 设置输出形状
box_results =box_results.reshape(8400,4)
conf_results = conf_results.reshape(8400,2)
scores = []
classes = []
boxes = []
for i in range(8400):
conf = conf_results[i,:] # 预测分数
score = np.max(conf) # 获取类别
# 筛选较小的预测类别
if score > 0.5:
classes.append(np.argmax(conf))
scores.append(score)
boxes.append(box_results[i,:])
scores = np.array(scores)
boxes = np.array(boxes)
result_box = []
result_score = []
result_class = []
# 非极大值抑制筛选重复的预测结果
if len(boxes) != 0:
# 非极大值抑制结果
indexs = tf.image.non_max_suppression(boxes,scores,len(scores),0.25,0.35)
for i, index in enumerate(indexs):
result_score.append(scores[index])
result_box.append(boxes[index,:])
result_class.append(classes[index])
# 返回结果
return np.array(result_box),np.array(result_score),np.array(result_class)
画出预测框
def draw_box(image, boxes, scores, classes, labels):
"""将预测结果绘制到图像上
Args:
image (uint8): 原图片
boxes (float32): 预测框
scores (float32): 分数
classes (int): 类别
lables (str): 标签
Returns:
uint8: 标注好的图片
"""
colors = [(0, 0, 255), (0, 255, 0)]
scale = max(image.shape) / 640.0 # 缩放比例
if len(classes) != 0:
for i in range(len(classes)):
box = boxes[i,:]
x1 = int(box[0] * scale)
y1 = int(box[1] * scale)
x2 = int(box[2] * scale)
y2 = int(box[3] * scale)
label = labels[classes[i]]
score = scores[i]
cv.rectangle(image, (x1, y1), (x2, y2), colors[classes[i]], 2, cv.LINE_8)
cv.putText(image,label+":"+str(score),(x1,y1-10),cv.FONT_HERSHEY_SIMPLEX, 0.55, colors[classes[i]], 2)
return image
读取标签
def read_label(label_path):
with open(label_path, 'r') as f:
labels = f.read().split()
return labels
同步推理
label_path = "labels.txt"
yoloe_model_path = "ppyoloe_crn_s_80.xml"
predictor = Predictor(model_path = yoloe_model_path)
boxes_name = predictor.get_outputs_name(0)
conf_name = predictor.get_outputs_name(1)
labels = read_label(label_path=label_path)
cap = cv.VideoCapture(0)
while cap.isOpened():
ret, frame = cap.read()
frame = cv.flip(frame, 180)
cv.namedWindow("MaskDetection", 0) # 0可调大小,注意:窗口名必须imshow里面的一窗口名一直
cv.resizeWindow("MaskDetection", 640, 480) # 设置长和宽
input_frame = process_image(frame, 640)
results = predictor.predict(input_data=input_frame)
boxes, scores, classes = process_result(box_results=results[boxes_name], conf_results=results[conf_name])
result_frame = draw_box(image=frame, boxes=boxes, scores=scores, classes=classes, labels=labels)
cv.imshow('MaskDetection', result_frame)
key = cv.waitKey(1)
if key == 27: #esc退出
break
cap.release()
cv.destroyAllWindows()
异步推理
label_path = "labels.txt"
yoloe_model_path = "ppyoloe_crn_s_80.xml"
predictor = Predictor(model_path = yoloe_model_path)
input_layer = predictor.get_inputs_name(0)
labels = read_label(label_path=label_path)
cap = cv.VideoCapture(0)
curr_request = predictor.get_request()
next_request = predictor.get_request()
ret, frame = cap.read()
curr_frame = process_image(frame, 640)
curr_request.set_tensor(input_layer, ov.Tensor(curr_frame))
curr_request.start_async()
while cap.isOpened():
ret, next_frame = cap.read()
next_frame = cv.flip(next_frame, 180)
cv.namedWindow("MaskDetection", 0) # 0可调大小,注意:窗口名必须imshow里面的一窗口名一直
cv.resizeWindow("MaskDetection", 640, 480) # 设置长和宽
in_frame = process_image(next_frame, 640)
next_request.set_tensor(input_layer, ov.Tensor(in_frame))
next_request.start_async()
if curr_request.wait_for(-1) == 1:
boxes_name = curr_request.get_output_tensor(0).data
conf_name = curr_request.get_output_tensor(1).data
boxes, scores, classes = process_result(box_results=boxes_name, conf_results=conf_name)
frame = draw_box(image=frame, boxes=boxes, scores=scores, classes=classes, labels=labels)
cv.imshow('MaskDetection', frame)
frame = next_frame
curr_request, next_request = next_request, curr_request
key = cv.waitKey(1)
if key == 27: #esc退出
break
cap.release()
cv.destroyAllWindows()
推理代码中,除了同步推理和异步推理,其他几个部分代码参考于这篇文章基于OpenVINO部署PaddlePadle-YOLOE模型—4.Python实现,并对代码进行修改。
口罩检测
点击此处获取代码
推理结果显示,用Intel CPU很丝滑,在OpenVINO 模型推理器类中,device_name除了可以使用Intel的CPU、GPU、VPU还有FPGA。所以接下来,你懂的,我们准备把这个训练好的模型迁移到FPGA上,看看有什么不一样的奇迹。至此,口罩检测就结束了。我们会在哪里再一次相遇呢?see you!