Question

当我尝试为keras创建自定义指标时，我遇到了错误（Intersection over union）。我想找到两个图像（张量）联合的交叉点

def IoU(y_true,y_pred):
    y_true_f = K.flatten(y_true)
    y_pred_f = K.flatten(y_pred)
    #assert len(y_true_f) != len(y_pred_f)
    y_true_f = y_true_f.eval(session = K.get_session())
    y_pred_f = y_pred_f.eval(session = K.get_session())
    union1 = [i  for i,j in zip(y_true_f,y_pred_f) if i != j]
    union2 = [j  for i,j in zip(y_true_f,y_pred_f) if i != j]
    intersection = [i for i,j in zip(y_true_f,y_pred_f) if i == j]
    unionAll = union1 + union2 + intersection
    return (np.sum(intersection) + smooth) / float(np.sum(unionAll)+ smooth)

错误我得到了：

InvalidArgumentError（请参阅上面的回溯）：您必须提供值对于占位符张量＆activation = activation_1_target＆＃39;与dtype浮动和形状[？，？，？] [[节点：activation_1_target = Placeholderdtype = DT_FLOAT，shape = [？，？，？]， _device =＆＃34; / job：localhost / replica：0 / task：0 / gpu：0＆＃34;]] [[Node：metrics / IoU / Reshape / _5 = _Recvclient_terminated = false， recv_device =＆＃34; /作业：本地主机/复制：0 /任务：0 / CPU：0＆＃34 ;, send_device =＆＃34; /作业：本地主机/复制：0 /任务：0 / GPU：0＆＃34 ;, send_device_incarnation = 1，tensor_name =＆＃34; edge_8_metrics / IoU / Reshape＆＃34;， tensor_type = DT_FLOAT， _device =＆＃34; /作业：本地主机/复制：0 /任务：0 / CPU：0＆＃34;]]

Answer 1

eps = 1.
def iou(y_true, y_pred):
   y_true_f = K.flatten(y_true)
   y_pred_f = K.flatten(y_pred)
   intersection = K.sum(y_true_f*y_pred_f)
   union = K.sum(y_true_f)+K.sum(y_pred_f)-intersection+eps
   return intersection/union

IoU keras（后端tenserflow）为占位符错误（tensor.eval（）提供dict

1 个答案: