YOLOv2 'ValidationData' error

Hello!

I'm trying to run the YOLOv2 example - https://www.mathworks.com/help/deeplearning/ug/object-detection-using-yolo-v2.html?s_tid=blogs_rc_6

However each time i see the error:

Error using trainingOptions (line 285)
The value of 'ValidationData' is invalid. The datastore used for 'ValidationData' must return a 2-column table or an
M-by-2 cell array.

Error in YOLOv3 (line 97)
options = trainingOptions('sgdm',...

Could you give me an advice how to solve the problem? The source is the same as the example...

[Marco Costanzo]

I have the same problem, any solution?

Answer 1

[Miles Brim]

I reached out to MATLAB and got the following response. I had the same error and this resolved it.

It appears that the issue you are encountering is due to changes that have been made to the Computer Vision Toolbox between R2019b and R2020a.

The 'Object Detection Using YOLO v2 Deep Learning' site that you included a link to in the original case description is specific to R2020a. The R2019b version can be found here:

https://www.mathworks.com/help/releases/R2019b/deeplearning/examples/object-detection-using-yolo-v2.html

There are quite a few lines of code that differ between the two versions, so I would recommend creating a new script file and copying the entirety of the R2019b example code.

In general, to ensure that you are using an example or function that is compatible with your current version of MATLAB, you can search for the example or function in your version’s archived documentation, found here:

https://www.mathworks.com/help/doc-archives.html