|
????????前言：本人最近在学着使用yolo，为了方便调用，今天看了下detect.py源码并做了点修改，不过由于训练的模型是单类型的，所以在推理在结果中还没看出哪个数值代表“类型”，所以以下内容只针对单类型的模型，后面琢磨出来了会更新。希望有大佬能指导一下。
在detect.py中添加如下代码块
class yolo_detector:
def __init__(self,
weights='./Weights/last.pt', # 用train.py训练出的.pt文件
imgsz=(640,640),
conf_thres=0.25,
iou_thres=0.45,
half=False,
):
self.conf_thres = conf_thres
self.iou_thres = iou_thres
self.device = select_device('0')
self.model = DetectMultiBackend(weights, device=self.device) # 加载模型
stride, names, pt = self.model.stride, self.model.names, self.model.pt
self.imgsz = check_img_size(imgsz, s=stride) # check image size
half &= pt and self.device.type != 'cpu' # half precision only supported by PyTorch on CUDA
self.half = half
if pt:
self.model.model.half() if half else self.model.model.float()
self.view_img = check_imshow()
cudnn.benchmark = True # set True to speed up constant image size inference
self.model.warmup(imgsz=(1, 3, *self.imgsz), half=self.half)
def run(self, frame):
# (h, w, c) to (c, h, w)
b, g, r = cv2.split(frame)
im0 = numpy.array([b, g, r])
im = torch.from_numpy(im0).to(self.device)
im = im.half() if self.half else im.float() # uint8 to fp16/32
im /= 255 # 0 - 255 to 0.0 - 1.0
if len(im.shape) == 3:
im = im[None] # expand for batch dim
pred = self.model(im)
pred = non_max_suppression(pred, self.conf_thres, self.iou_thres, max_det=5)
results = []
for i, det in enumerate(pred):
if len(det):
# Rescale boxes from img_size to im0 size
det[:, :4] = scale_coords(im.shape[2:], det[:, :4], frame.shape).round()
if det.numel():
x1, y1, x2, y2 = int(det[0, 0].item()), int(det[0, 1].item()), int(det[0, 2].item()), int(det[0, 3].item())
lu = (x1, y1)
rd = (x2, y2)
results.append((lu, rd))
return results
? ? ? ?调用实例：
import cv2
import detect_remake
cap=cv2.VideoCapture(0)
a = detect_remake.yolo_detector()
while True:
rec,img = cap.read()
results = a.run(img)
if results:
for i, pts in enumerate(results):
cv2.rectangle(img, pts[0], pts[1], (0, 0, 255), 2)
cv2.imshow("video",img)
if cv2.waitKey(1)==ord('q'):
break
|