Error: could not broadcast input array from shape (394,394,3) into shape (200,200,3)

Data Prep and Ingestion
#1

I get this error when in the training window, when i click “Start Training”:

could not broadcast input array from shape (394,394,3) into shape (200,200,3)

There is no “!” in the model tab, so the dataset+ model is supposed to be fine. I used autoML. I left the hyperparameters at default. My train.csv looks like this:
image

How can I fix this problem?

#2

Hi @phzHRO,

Can you try this command on your image dataset:

identify -format "%i %[colorspace]\n" *.jpg
and check if there are images of other formats and also ensure to have the images of same size, else use resize option provided by DLS.

1 Like
#3

Thank you,
I had to activate the resize option in DLS, as I have images of different size. I did not know yet, that DLS can’t deal with this automatically.

Just out of curiosity, where exactly did you want me to insert your command?

#4

That’s a linux command to identify format and characteristics of one or more image files. So, i wanted you to check if there is any other format in your dataset other than JPG, before you had uploaded your dataset.

#5

Okay. I didn’t know that command. I’m on windows.

#6

So t he DLS just supports jpg? well then this could be the reason for my

“could not broadcast input array from shape (200,200,4) into shape (200,200,3)”