I’ve written Lesson 2 on Google Colab, you can check it out here. As instructed in the documentation I’ve mounted Google Drive to the Colab instance as well and downloaded a couple of “bear” and “pandas” images using the duckgoose library.
When I try to verify the dataset for “good” images, I’m prompted with this error; NameError: name 'classes' is not defined
. I can’t figure out what mistake did I make, although I’ve a feeling that since duckgoose
automatically makes train, test and validation sets, fastai
can’t figure out the labels for the image dataset.
EDIT: I believe I’ve found out what went wrong here. I’m not using the download_image()
function as such there’s no way for me to verify the integrity of the images. Is there an alternative for the same thing? I would really like to verify the images before running it through a neural network.
I had this following changes and it works fine!
classes = [' bear', 'pandas']
path = Path('data/ducksgeese')
for c in classes:
print(c)
verify_images(path/c, delete=True, max_workers=8)