The dataset is small. It contains the following 20 categories:
- bike
- crab
- ipod
- license-plate
- owl
- playing-card
- raccoon
- smokestack
- spaghetti
- syringe
- chandelier
- grapes
- ketch
- octopus
- paperclip
- pyramid
- skateboard
- soda-can
- speed-boat
- umbrella
The training data contains the above-mentioned 20 categories, with 40 images in each category (800 images total).