input layer and flatten mismatch

2021 12 12
1 个回答
回答已采纳
更新时间：2021 12 12
9 次查看（30 天）

 采纳的回答

0 个投票

The Flatten layer support sequence input only. [1]
As i understand from your network architechture you don't need Flatten Layer. in Matlab Flattenlayer is for merging height,width and channel of input in an aray for N data in minibatch and S (number of sequences) series data. the output is a 3-dimensianal array with size of (H*W*C)xNxS.
if you use a fullyconnectedlayer(fc) as you did after Flatten layer, the input array of fc layer which is comming from a Conv layer become vectorize in a spatial domain in output of layer.
so just deleting the Flatten layer will solve the problem.
[1] https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.flattenlayer.html

更多回答（0 个）

类别

在 帮助中心 和 File Exchange 中查找有关 Deep Learning Toolbox 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by