數(shù)據(jù)集地址

該圖像數(shù)據(jù)集包含8000張圖像，兩個(gè)類別分別是安全帽與人、以其中200多張圖像為驗(yàn)證集，其余為訓(xùn)練集。

模型訓(xùn)練

準(zhǔn)備好數(shù)據(jù)集以后，直接按下面的命令行運(yùn)行即可：

yolotrainmodel=yolov8s.ptdata=hat_dataset.yamlepochs=50imgsz=640batch=4

導(dǎo)出與測(cè)試

模型導(dǎo)出與測(cè)試

yolo export model=hat_best.pt format=onnx
yolo predict model=hat_best.pt source=./hats

部署推理

轉(zhuǎn)成ONNX格式文件以后，基于OpenVINO-Python部署推理，相關(guān)代碼如下

#ReadIR
model=ie.read_model(model="hat_best.onnx")
compiled_model=ie.compile_model(model=model,device_name="CPU")
output_layer=compiled_model.output(0)

capture=cv.VideoCapture("D:/images/video/hat_test.mp4")
whileTrue:
_,frame=capture.read()
ifframeisNone:
print("Endofstream")
break
bgr=format_yolov8(frame)
img_h,img_w,img_c=bgr.shape

start=time.time()
image=cv.dnn.blobFromImage(bgr,1/255.0,(640,640),swapRB=True,crop=False)

res=compiled_model([image])[output_layer]#1x84x8400
rows=np.squeeze(res,0).T
class_ids=[]
confidences=[]
boxes=[]
x_factor=img_w/640
y_factor=img_h/640

forrinrange(rows.shape[0]):
row=rows[r]
classes_scores=row[4:]
_,_,_,max_indx=cv.minMaxLoc(classes_scores)
class_id=max_indx[1]
if(classes_scores[class_id]>.25):
confidences.append(classes_scores[class_id])
class_ids.append(class_id)
x,y,w,h=row[0].item(),row[1].item(),row[2].item(),row[3].item()
left=int((x-0.5*w)*x_factor)
top=int((y-0.5*h)*y_factor)
width=int(w*x_factor)
height=int(h*y_factor)
box=np.array([left,top,width,height])
boxes.append(box)

indexes=cv.dnn.NMSBoxes(boxes,confidences,0.25,0.45)
forindexinindexes:
box=boxes[index]
color=colors[int(class_ids[index])%len(colors)]
cv.rectangle(frame,box,color,2)
cv.rectangle(frame,(box[0],box[1]-20),(box[0]+box[2],box[1]),color,-1)
cv.putText(frame,class_list[class_ids[index]],(box[0],box[1]-10),cv.FONT_HERSHEY_SIMPLEX,.5,(0,0,0))
end=time.time()
inf_end=end-start
fps=1/inf_end
fps_label="FPS:%.2f"%fps
cv.putText(frame,fps_label,(20,45),cv.FONT_HERSHEY_SIMPLEX,1,(0,0,255),2)

cv.imshow("YOLOv8hatDetection",frame)
cc=cv.waitKey(1)
ifcc==27:
break
cv.destroyAllWindows()