Deep Learning Keras preprocessing node won't configure

users
#1

I can’t configure the Load and pre-process images node in the 01-Preprocess_image_data example Deep Learning - Keras workflow. The node image is blank, and it issues a fatal error when I try to execute it. The configuration screen says that there is no Quick Form available. I think that I installed all the proper extensions, I installed and activated a conda environment in Anaconda Python 3.7, and downloaded the cats and dogs data folders.
What did I do wrong?

Bob Nisbet

0 Likes

#2

Hi @dataminer_1,

In order to configure this Wrapped Metanode (learn more about them here) you would have to properly configure and execute the List Files node at the very beginning of this workflow. To properly configure it, please follow the instructions detailed in the workflow annotation at the very top, where the first step is to download and unzip a specific file.

This will, however, not enable configuration of the Wrapped Metanode. If you want to configure individual nodes, you can open it by Ctrl/Cmd and double click on the icon.

Best regards,
Stefan

1 Like

#3

I did that. The cats-vs-dogs folder I downloaded has only 4,000 dogs and 4,000 cats. I loaded the folder successfully into the List Files node and reconfigured the Rule engine to specify that the first 4000 rows were cats (instead of 125,000). I executed the Rules engine node successfully, showing a RowID and 3 data columns. Now, I can’t execute the Load and Preprocess node successfully. In the Console, I get the error: 0:241:0:252 Execute failed: Cell count in row “Row0” is not equal to length of column names array: 1 vs. 3.

Bob

0 Likes

#4

I have just downloaded the dogs-vs-cats data and unzipping the train.zip file into a folder results in 25000 files. You could try to re-download the data since something might have gone wrong during the first download?

While the error message is a little misleading, this hints at corrupt image files that cannot be read. I would suggest to re-download the files and give it another try…

3 Likes

#5

Stelfrich:

Thanks for the guidance. I think that I downloaded the wrong version of the dogs-vs-cats data set. I downloaded the 25K data set from Kaggle, and it worked fine.
Thanks your help!
Bob

1 Like