error using classify on a trained cnn: For an image input layer, the input data must be a single image, a 4-D array of images, or an imageDatastore with the correct size.

Question

MathWorks Support Team 2018-4-6

2
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/395583-error-using-classify-on-a-trained-cnn-for-an-image-input-layer-the-input-data-must-be-a-single-ima

编辑： MathWorks Support Team 2021-9-27

采纳的回答： MathWorks Support Team

在 MATLAB Online 中打开

I have trained a cnn using "trainNetwork" following the workflow outlined in the "trainNetwork" documentation page:

https://www.mathworks.com/help/deeplearning/ref/trainnetwork.html

I would like to now use my trained network to classify images via the function "classify" (also mentioned in the example in the documentation.) When I call "classify", I get the following error:

ERROR: Error using SeriesNetwork/predict (line 331)
For an image input layer, the input data must be a single image, a 4-D array of images, or an imageDatastore with the correct size.
Error in SeriesNetwork/classify (line 561)
            scores = this.predict( X, varargin{:} );

However, I have checked my inputs to "classify" and ensured that the first input is a SeriesNetwork and the second is an ImageDatastore. What am I doing wrong?

请先登录，再回答此问题。

Answer 1

MathWorks Support Team 2021-8-17

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/395583-error-using-classify-on-a-trained-cnn-for-an-image-input-layer-the-input-data-must-be-a-single-ima#answer_315659

编辑：MathWorks Support Team 2021-9-27

As per the error, make sure that imageDatastore contains images with the correct size (pertaining to the sizes of the images used to train the network.) If the sizes do not match up, "imresize" can be used to adjust the sizes of the images to be classified.