A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets
Download and extract dataset:
python utils/prepare_dataset.py --dataset SmallImageNet --resolution 32 --data-dir data --download-dir data/compressed
Supported resolutions: 8, 16, 32, 64 (must be >=32 for ImageNet ResNets)
Training:
CIFAR ResNets: python train.py data --dataset SmallImageNet --size 32 --classes 1000 --depth 20 --ngpu 1 --epochs 200 -b 128 --lr 0.1 --momentum 0.9 --wd 5e-4 --prefix test --project Imagenet
ImageNet ResNets: python train.py data --dataset SmallImageNet --size 32 --classes 1000 --depth 18 --ngpu 1 --epochs 100 -b 256 --lr 0.1 --momentum 0.9 --wd 1e-4 --prefix test --project Imagenet
classes
can be changed to select a subset of the dataset. size
is the resolution of the dataset.
Tiny ImageNet Visual Recognition Challenge
Download and extract dataset:
python utils/prepare_dataset.py --dataset TinyImageNet --data-dir data --download-dir data/compressed
Training:
CIFAR ResNets: python train.py data --dataset TinyImageNet --depth 20 --ngpu 1 --epochs 200 -b 128 --lr 0.1 --momentum 0.9 --wd 5e-4 --prefix test --project Imagenet
ImageNet ResNets: python train.py data --dataset TinyImageNet --depth 18 --ngpu 1 --epochs 100 -b 256 --lr 0.1 --momentum 0.9 --wd 1e-4 --prefix test --project Imagenet
Run python utils/compute_stats.py --dataset SmallImageNet --resolution 8 --data-dir data
to compute mean and std. of the dataset. dataset_stats.json
contains stats for 1000 classes.