search

truc-h-nguyen / Toddler-activity-suggestions

1 stars 0 forks source link

Save dataset #7

Closed truc-h-nguyen closed 2 years ago

truc-h-nguyen commented 2 years ago

@nickvazz Hi Nick, In [16], I couldn't download another dataset if the "data" folder is already occupied.

The "bowl" set has about 7k entries while the "teddy bear" set has about 2k. I take number of samples for "bowl" train set is 2000. Can I take 1000 samples for "teddy bear" train set or the number of samples of both train sets need to be the same?

nickvazz commented 2 years ago

Ideally you have the same number of samples for each class, but there are ways to deal with an imbalanced dataset