Can I use the model in the lesson 1 to do multi-classification jobs? If yes, where can I specify the number of classes to be classified?
In addition, if I use ImageClassifierData.from_arrays, it keeps saying “NotImplementedError: uint8”, but my images’ shape was exactly same as the images in “dogscats” files. Could anyone know the problem? Thank you very much.
Yes, you can use it to do multi classification jobs (assuming each input can belong to only one of the many labels). This should work as long as you have put the data for multiple labels in their own separate folders respectively(for data in train and valid folders).
If you want to do multi classification, look at the planet example notebook. It uses a CSV with the image id and a space separated list of labels per row. All images are in a single directory.