与输入维有关的多输入Keras模型出现错误 [英] Getting an error with a multi-input Keras model related to the input dimension

问题描述

我有一个多输入Keras模型。这里是输入：

I have a multi input Keras model. Here the inputs:

[<tf.Tensor 'input_1:0' shape=(None, 256, 256, 3) dtype=float32>,
 <tf.Tensor 'input_2:0' shape=(None, 256, 256, 3) dtype=float32>,
 <tf.Tensor 'input_3:0' shape=(None, 256, 256, 3) dtype=float32>,
 <tf.Tensor 'input_4:0' shape=(None, 256, 256, 3) dtype=float32>]

这里是模型的输入形状：

And here the input shape of the model :

[(None, 256, 256, 3),
 (None, 256, 256, 3),
 (None, 256, 256, 3),
 (None, 256, 256, 3)]

训练数据形状如下：

(4, 422, 256, 256, 3)
4 = number of inputs (consist of appended arrays together).
422 = number of training images in each input.
256, 256, 3 = shape of the images

当我称 fit 函数：

model.fit(train_x, train_y, validation_split=0.20, epochs=5, batch_size=3)

发生以下错误：

ValueError：输入层conv1_pad_0的0与该层不兼容：预期的ndim = 4，找到的ndim = 5。收到的完整图形：[3，422，256，256，3]

ValueError: Input 0 of layer conv1_pad_0 is incompatible with the layer: expected ndim=4, found ndim=5. Full shape received: [3, 422, 256, 256, 3]

我已经尝试过这篇文章，但我的基数不匹配。

I have tried the solution given in this post, but I got a mismatch in cardinality.

ValueError：数据基数模棱两可：

ValueError: Data cardinality is ambiguous:

我曾尝试像下面这样传递火车数据，并且有效：

I have tried passing the train data like bellow and it worked:

model.fit([train_x[0], train_x[1], train_x[2], train_x[3]], train_y, validation_split=0.20, epochs=5, batch_size=3)

现在，如果我想将模型缩放到20个输入，则上面的代码行会出现问题。

Now If I want to scale my model to 20 inputs the above line of code will be problematique.

更新：

该模型基于预先训练的 ResNet50 ，所有输入均为一个没有顶层的resnet50，并从以下三个层开始：

The model are based on the pretrained ResNet50, all inputs are a resnet50 without the top layers and begin with the following three layers :

input_1_0 (InputLayer)        [(None, 256, 256, 3) 0  
conv1_pad_0 (ZeroPadding2D)   (None, 262, 262, 3)  0           input_1_0[0][0]
conv1_conv_0 (Conv2D)         (None, 128, 128, 64) 9472        conv1_pad_0[0][0]   

用于训练/测试模型的数据按以下方式处理：

The data for training/testing the model is processed as follows:

for row in np.array(tmp_data):
        row = images_preprocessing(row) # Depends on the model used
        train_x, test_x, train_y, test_y = split_data(row, target) # Here the train_test_split is used
        
        train_X.append(train_x)
        test_X.append(test_x)
        train_Y.append(train_y)
        test_Y.append(test_y)

推荐答案

尝试

train_x_list = [tf.squeeze(tx) for tx in tf.split(train_x, num_or_size_splits=train_x.shape[0], axis=0)]

它将生成张量的列表，其中训练数据沿维度0拆分。然后使用第二个解决方案，将列表馈送到 fit（）。

it will produce a list of tensors with training data split along dimension 0. Then use your second solution, feeding the list to fit().

这篇关于与输入维有关的多输入Keras模型出现错误的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！