def detect(self, img_path):
img = transforms.ToTensor()(Image.open(img_path).convert('RGB'))
_, H, W = img.shape
h_factor, w_factor = (H, W)
img, pad = pad_to_square(img, 0)
_, padded_h, padded_w = img.shape
imgs = torch.stack([resize(img_, 416) for img_ in [img]]).to('cuda')
with torch.no_grad():
output = self.net(imgs)
output = non_max_suppression(output, 0.5, 0.5)[0]
output = rescale_boxes(output, 416, (H, W)).numpy()
detections = []
boxes = []
confidences = []
class_ids = []
for x1, y1, x2, y2, conf, cls_conf, cls_pred in output:
width = x2 - x1
height = y2 - y1
x = x1
y = y1
boxes.append([x, y, int(width), int(height)])
confidences.append(float(cls_conf))
class_ids.append(int(cls_pred))
class_ = self.get_class(int(cls_pred))
top_left = (int(x1), int(y1))
bottom_right = (int(x2), int(y2))
box_2d = [top_left, bottom_right]
detections.append(Detection(box_2d, class_))
return detections
def predict(self,
image,
device='cuda',
img_size=446,
conf_thres=0.6,
nms_thres=0.2,
normalize=False,
normalize_input=True):
"""
Util method for getting preprocessed bboxes
img - (W,H,3)
returns - bboxes (N, 7)
"""
if normalize_input:
img = transforms.ToTensor()(image)
else:
img = torch.from_numpy(np.array(image)).float().permute(2, 0, 1)
padded_img, pad = pad_to_square(img)
warped_img = F.interpolate(padded_img.unsqueeze(0),
size=img_size,
mode="nearest").squeeze(0)
if device:
device = torch.device(device)
self = self.to(device)
warped_img = warped_img.to(device)
result = self(warped_img.unsqueeze(0))
outputs = non_max_suppression(result,
conf_thres=conf_thres,
nms_thres=nms_thres)
pred_bboxes = outputs[0]
if pred_bboxes is None:
return []
_, img_h, img_w = img.shape
_, width, height = warped_img.shape
# normalize back
pred_bboxes[:, [0, 2]] /= width
pred_bboxes[:, [1, 3]] /= height
pred_bboxes[:, [0, 2]] -= pad[0] / max(img_w, img_h)
pred_bboxes[:, [1, 3]] -= pad[2] / max(img_w, img_h)
if not normalize:
pred_bboxes[:, [0, 2]] *= max(img_w, img_h)
pred_bboxes[:, [1, 3]] *= max(img_w, img_h)
return pred_bboxes
def infer_video(model,
video_path,
save_path,
iou_thres,
conf_thres,
nms_thres,
img_size,
batch_size=1):
model.eval()
if os.path.isfile(video_path):
cap = cv2.VideoCapture(video_path)
else:
raise Exception('no such video')
tp = time.strftime('%Y-%m-%d-%H-%M-%S', time.localtime())
stream = cv2.VideoWriter(os.path.join(save_path, 'out.mp4'),
cv2.VideoWriter_fourcc(*'MJPG'), 20.0,
(int(cap.get(3)), int(cap.get(4))))
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
while cap.isOpened():
ret, frame = cap.read()
if ret:
print(frame.shape)
img = transform(frame).to(device)
print(img.shape)
img, _ = pad_to_square(img, 0)
print(img.shape)
img = F.interpolate(img.unsqueeze(0), img_size,
mode='nearest').squeeze()
print(img.shape)
img = img.unsqueeze(0)
with torch.no_grad():
output = model(img)
output = non_max_suppression(output,
conf_thres=conf_thres,
nms_thres=nms_thres)
output = rescale_boxes(output[0], img_size, frame.shape[:2])
# print(output)
# print(len(output[0][0]))
img = vis_frame(frame, output)
stream.write(img)
else:
break
cap.release()
stream.release()
def Yolo_detect(model,
camInputFrame,
img_size=416,
conf_thres=0.8,
nms_thres=0.4):
img = transforms.ToTensor()(Image.fromarray(camInputFrame))
# Pad to square resolution
img, _ = pad_to_square(img, 0)
# Resize
img = resize(img, img_size)
img = img.unsqueeze(0) #(1,3,416.419)
input_imgs = img.cuda()
with torch.no_grad():
detections = model(input_imgs)
detections = non_max_suppression(detections, conf_thres, nms_thres)
if detections is not None:
detections = detections[0]
if detections is not None:
detections = rescale_boxes(detections, img_size,
camInputFrame.shape[:2])
return detections
ret, frame = input.read()
nums += 1
if not ret:
break
print("---------------读取第" + str(nums) + "帧")
frame_start_t = time.time()
# cv2 读取图片转换为 PIL 格式 转换为 Tensor
# img = torchvision.transforms.ToTensor()(Image.open(img_path).convert(mode="RGB"))
frame_pil = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
img = torchvision.transforms.ToTensor()(frame_pil.convert(mode="RGB"))
# NEW 创建一个可用来对其进行draw的对象
draw = ImageDraw.Draw(frame_pil)
input_imgs, _ = pad_to_square(img, 0)
# Resize
input_imgs = resize(input_imgs, opt.img_size).unsqueeze(0)
# Configure input
input_imgs = Variable(input_imgs.type(Tensor))
tensor_t = time.time()
print("转换为 Tensor 用时：" + str(time.time() - frame_start_t))
# 开始检测
with torch.no_grad():
detections = model(input_imgs.to(device))
detections = non_max_suppression(detections, opt.conf_thres,
opt.nms_thres)[0]
detect_t = time.time()
print("进行物体检测用时：" + str(detect_t - tensor_t))
20.0, (int(cap.get(3)), int(cap.get(4))) )
trans = transforms.Compose([
transforms.ToTensor(),
# transforms.Normalize(mean=[0.486, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])
plt.figure()
plt.margins(0, 0)
while cap.isOpened():
ret, frame = cap.read()
if ret:
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame = Image.fromarray(frame)
img = trans(frame)
img, _ = pad_to_square(img, 0)
img = F.interpolate(img.unsqueeze(0), opt.img_size, mode='bilinear').squeeze()
img = img.unsqueeze(0).to(device)
with torch.no_grad():
detections = model(img)
detections = non_max_suppression(detections, opt.conf_thres, opt.nms_thres)
# Bounding-box colors
cmap = plt.get_cmap("tab20b")
colors = [cmap(i) for i in np.linspace(0, 1, 20)]
# Create plot
# img = np.array(frame)
# plt.figure()
fig, ax = plt.subplots(1)
ax.imshow(np.array(frame))
# Draw bounding boxes and labels of detections
def test_dataloader_getitem(self):
unloader = transforms.ToPILImage()
self.normalized_labels = True
# img_path = 'e:/ML_data/images/train2014/COCO_train2014_000000000092.jpg'
# label_path = 'e:/ML_data/labels/train2014/COCO_train2014_000000000092.txt'
img_path = 'E:/trafficlight_detect/light_img/IMG_20181124_125821.jpg'
label_path = 'E:/trafficlight_detect/labels/IMG_20181124_125821.txt'
# Extract image as PyTorch tensor
img = transforms.ToTensor()(Image.open(img_path).convert('RGB'))
fig, ax = plt.subplots(1)
image_orig = unloader(img)
ax.imshow(image_orig)
#ax.add_patch( patches.Rectangle( xy=( 225, 131), width=75, height=50, linewidth=1 ))
plt.show()
# Handle images with less than three channels
if len(img.shape) != 3:
img = img.unsqueeze(0)
img = img.expand((3, img.shape[1:]))
_, h, w = img.shape
h_factor, w_factor = (h, w) if self.normalized_labels else (1, 1)
img, pad = pad_to_square(img, 0)
_, padded_h, padded_w = img.shape
image_padded = unloader(img)
fig, ax = plt.subplots(1)
ax.imshow(image_padded)
plt.show()
boxes = torch.from_numpy(np.loadtxt(label_path).reshape(-1, 5))
x1 = w_factor * (boxes[:, 1] - boxes[:, 3] / 2)
y1 = h_factor * (boxes[:, 2] - boxes[:, 4] / 2)
x2 = w_factor * (boxes[:, 1] + boxes[:, 3] / 2)
y2 = h_factor * (boxes[:, 2] + boxes[:, 4] / 2)
fig, ax = plt.subplots(1)
ax.imshow(image_orig)
#ax.add_patch( patches.Rectangle( xy=( 225, 131), width=75, height=50, linewidth=1 ))
for box_i, (xx1, yy1, xx2, yy2) in enumerate(zip(x1, y1, x2, y2)):
ax.add_patch(
patches.Rectangle(xy=(xx1.item(), yy1.item()),
width=(xx2 - xx1).item(),
height=(yy2 - yy1).item(),
linewidth=1))
plt.show()
# Adjust for added padding
x1 += pad[0]
y1 += pad[2]
x2 += pad[1]
y2 += pad[3]
fig, ax = plt.subplots(1)
ax.imshow(image_padded)
#ax.add_patch( patches.Rectangle( xy=( 225, 131), width=75, height=50, linewidth=1 ))
for box_i, (xx1, yy1, xx2, yy2) in enumerate(zip(x1, y1, x2, y2)):
ax.add_patch(
patches.Rectangle(xy=(xx1.item(), yy1.item()),
width=(xx2 - xx1).item(),
height=(yy2 - yy1).item(),
linewidth=1))
plt.show()
boxes[:, 1] = ((x1 + x2) / 2) / padded_w
boxes[:, 2] = ((y1 + y2) / 2) / padded_h
boxes[:, 3] = boxes[:, 3] * w_factor / padded_w
boxes[:, 4] = boxes[:, 4] * h_factor / padded_h
fig, ax = plt.subplots(1)
ax.imshow(image_padded)
#ax.add_patch( patches.Rectangle( xy=( 225, 131), width=75, height=50, linewidth=1 ))
for box_i in range(len(boxes)):
x = boxes[box_i, 1] * padded_w
y = boxes[box_i, 2] * padded_h
width = boxes[box_i, 3] * padded_w
height = boxes[box_i, 3] * padded_h
ax.add_patch(
patches.Circle(xy=(x.item(), y.item()), radius=10,
linewidth=1))
plt.show()
def YOLO():
parser = argparse.ArgumentParser()
parser.add_argument("--model_def", type=str, default="config/yolov3.cfg", help="path to model definition file")
parser.add_argument("--weights_path", type=str, default="weights/yolov3.weights", help="path to weights file")
parser.add_argument("--class_path", type=str, default="data/coco.names", help="path to class label file")
parser.add_argument("--conf_thres", type=float, default=0.8, help="object confidence threshold")
parser.add_argument("--nms_thres", type=float, default=0.2, help="iou thresshold for non-maximum suppression")
parser.add_argument("--batch_size", type=int, default=1, help="size of the batches")
parser.add_argument("--img_size", type=int, default=416, help="size of each image dimension")
parser.add_argument("--video", type=str, required=True, help="input video")
parser.add_argument("--display", action="store_true", default=False)
parser.add_argument("--output", default="./output", help="output dir")
opt = parser.parse_args()
print(opt)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = Darknet(opt.model_def, img_size=opt.img_size).to(device)
if opt.weights_path.endswith(".weights"):
# Load darknet weights
model.load_darknet_weights(opt.weights_path)
else:
# Load checkpoint weights
model.load_state_dict(torch.load(opt.weights_path))
model.eval() # Set in evaluation mode
classes = load_classes(opt.class_path) # Extracts class labels from file
Tensor = torch.cuda.FloatTensor if torch.cuda.is_available() else torch.FloatTensor
#cap = cv2.VideoCapture(0)
cap = cv2.VideoCapture(opt.video)
cap.set(3, 1280)
cap.set(4, 720)
# out = cv2.VideoWriter(
# "output.avi", cv2.VideoWriter_fourcc(*"MJPG"), 10.0,
# (darknet.network_width(netMain), darknet.network_height(netMain)))
print("Starting the YOLO loop...")
while True:
try:
prev_time = time.time()
ret, frame_read = cap.read()
frame = cv2.cvtColor(frame_read, cv2.COLOR_BGR2RGB)
# Extract image as PyTorch tensor
img = transforms.ToTensor()(frame)
# Pad to square resolution
img, _ = pad_to_square(img, 0)
# Resize
img = resize(img, opt.img_size)
img = img.unsqueeze(0)
# Configure input
input_imgs = nn.Variable(img.type(Tensor))
# Get detections
with torch.no_grad():
detections = model(input_imgs)
detections = non_max_suppression(detections, opt.conf_thres, opt.nms_thres)
frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
detections = list(filter(lambda x: x is not None, detections))
if detections is not None and len(detections) > 0:
# Rescale boxes to original image
detections = rescale_boxes(detections[0], opt.img_size, frame.shape[:2])
frame = cvDrawBoxes(frame, detections, classes)
current_time = datetime.datetime.now()
if int(time.time()*10) % 10 == 0:
str_date = datetime.datetime.strftime(current_time, "%Y%m%d")
str_time = datetime.datetime.strftime(current_time, "%Y%m%d%H%M%S")
os.makedirs(os.path.join(opt.output, str_date), exist_ok=True)
cv2.imwrite(os.path.join(opt.output, str_date, str_time + ".jpg"), frame)
# print(1/(time.time()-prev_time))
if opt.display:
cv2.imshow('Demo', frame)
cv2.waitKey(3)
except Exception as e:
print("fail to detect", e)
cap.release()