我正在训练深度学习模型,但获得的准确性非常低,但损失也很低,这与二者成反比.这种低精度的原因是什么,我该如何制止它?
该模型应检测视频场景中的异常事件
我用来建立模型的研究论文的链接
模型架构
这是我的代码:
model = Sequential()
model.add(TimeDistributed(Convolution2D(512, 11, 11 ,activation='relu' , border_mode='valid', subsample = (4,4))
, input_shape=(10,231, 231, 1)))
model.add(TimeDistributed(MaxPooling2D(pool_size=(2, 2), strides=None, border_mode='valid')))
model.add(TimeDistributed(Convolution2D(256,5, 5,activation='relu' , border_mode='same')))
model.add(TimeDistributed(MaxPooling2D(pool_size=(2, 2), strides=None, border_mode='valid')))
model.add(TimeDistributed(Convolution2D(128,3,3,activation='relu' , border_mode='same'))),
model.add(TimeDistributed(Conv2DTranspose(128, 3,3,border_mode='same')))
model.add(TimeDistributed(UpSampling2D(size=(2, 2, ))))
model.add(TimeDistributed(Conv2DTranspose(256, 3,3,border_mode='same')))
model.add(TimeDistributed(UpSampling2D(size=(2, 2, ))))
model.add(TimeDistributed(Conv2DTranspose(512,5,5,border_mode='same')))
model.add(TimeDistributed(Conv2DTranspose(1, kernel_size=(11,11),strides=(4,4))))