We pull images from Tiny-Imagenet and create three datasets for Micro-robot purposes. Each class has 500 training images and 50 testing images:
10 classes of outdoor objects a micro-robot may encounter.
- acorn
- basketball
- bee
- black_widow
- frog
- ladybug
- mushroom
- nail
- snail
- tarantula
10 classes of indoor objects a micro-robot may encounter.
- banana
- candle
- dumbbell
- iPod
- pill_bottle
- pop_bottle
- sandal
- sock
- syringe
- teapot
20 classes combining the indoor and outdoor objects.
- Run
sh setup.sh
in the current directory. - Use an
ImageFolder
/DataLoader
in PyTorch, implementation from outside this directory.