orgtrack's Introduction

OrgTrack

直接接入模型预测接口即可，或直接使用延时标签图像进行单个类器官图像追踪

修改 main.py 中网络：

  if  model_name=="MACP":  
         """  
         网络预测接口  
         """  
         model = MACPNet.MACP.MVCnet()  
         model = model.to(device)  
         model.load_state_dict(torch.load(r'MACPNet\MACP.pth', map_location='cpu'))

或者使用网络分割好的mask进行stack后替换下面代码中的“imge”:

  if show_label==True:
      def img2bin(imge):
          newimg = []
          for i ,tempt in enumerate(imge) :
              # 示例输入张量
              input_tensor = tempt
              np_input=np.array(input_tensor)
              #边缘检测，hysteresisMaximum=0.05,hysteresisMinimum=0.005用于分离的高低阈值参数
              output_tensor_edge=(outtool.DetectEdges(np.expand_dims(np_input,axis=0),gaussianSigma=2.0,hysteresisMaximum=0.05,hysteresisMinimum=0.005,foregroundThreshold=0.5)).astype('int8')
              #分离，阈值foregroundThreshold=0.5
              output_tensor_seperate=outtool.SeparateContours(edges=output_tensor_edge,images=np.expand_dims(np_input,axis=0),foregroundThreshold=0.5,gaussianSigma=2)
              #填孔，过滤尺寸较小的类器官minimumArea=50
              output_tensor=outtool.Cleanup(output_tensor_seperate,minimumArea=50,fillHoles=True,removeBorders=False)
              # 遍历每张图像
              newimg.append(output_tensor[0])
          newimg = np.stack(newimg, axis=0)
          return newimg
      tailuo = np.expand_dims(img2bin(imge),axis=1)
      return tailuo,image_name

Recommend Projects