[Lesson 3] Multilabel from folder

Part 1 (2019)
1

Hello!

I tried to create my own multilabel model based on this dataset from Kaggle: https://www.kaggle.com/jutrera/stanford-car-dataset-by-classes-folder

The issue I’m facing is that the labels are in the folder name, such as “AM General Hummer SUV 2000”. Using the ImageDataBunch.from_folder is creating the title as one single label, and there is no label_delim parameter I could use to split the folder name.

Do you have any idea how to handle this kind of labeling?

Thank you,

Sebastien De Spiegeleer

2

Hi Sebastian,

I’m on my mobile, and can’t check the code and available functions right now. Thinking aloud, I would probably try to parse the folder names (using string substitution and/or RegEx) into a csv file with the image filenames and multi-labels, and then feed the data into fastai library that way.

Yijin