Select to view content in your preferred language

Omitting input classes in Training Deep Learning Model Tool

301
1
02-25-2024 02:07 PM
Bartek_Sobieszek
New Contributor

Hello everyone
i created many samples for many different camera trap images. The samples are the same - I marked the penguins. Unfortunately, each photo is called a different class (input data in Train Deep Learning Model tool is different for each folder)

How can I skip classes in sample training? I would like the tool to only take into account samples without class assignment.

I tried using this argument: "ignore_classes - list of class values for which the model will not suffer losses." but I get an error every time

ArcGIS Pro 3.2.2

BartomiejSobieszek_0-1708898730093.png

 

0 Kudos
1 Reply
PriyankaTuteja
Esri Contributor

Hello @Bartek_Sobieszek ,

 

Could you share what is the class value and class name within each folder of your training data? is it same or different? Since you have labelled penguins across all the folders. 

ignore_classes won't work as it is designed to ignore certain classes while training when the number of classes are more than 2. 

After understanding issue with your data classes, we could provide you a script that fixes your training data.

0 Kudos