How would you train a CNN to accurately predict the output of say the swt2 or fft2 function on an image?

1 次查看（过去 30 天）
Drew Ellington
Drew Ellington 2022-2-4
回答： yanqi liu 2022-2-7
Does anyone have any documentation, theories, code, methods or achitecture to train a CNN to accurately predict the output of say the swt2 or fft2 or any wavelet transform for a given image?
I'd be looking to train the CNN with a dataset of images as the X and the pre-processed images (swt2(ImageX)) as the Y value to predict an unknown image's wavelet transform output with high accuracy,
I'm just looking to gain an understanding of neural network's for less cookie-cutter problems but I'm struggling to find much documentation,
any responses are appreciated,
thanks.

请先登录，再回答此问题。

采纳的回答

yanqi liu
yanqi liu 2022-2-7
yes，sir，may be make the input as vector，and use convolution2dLayer to get one dimension data，then define cnn model，such as
imageInputLayer([vector_length 1 1])
convolution2dLayer([100 1],3,'Stride',1)
if possible, may be upload your data and preprocess method.

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Image Data Workflows 的更多信息

标签

产品


版本

R2021b

Community Treasure Hunt

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

Start Hunting!

Translated by