algorithm for making the Extremely Biased and Poorly Categorized (EBPC) dataset
- Combine the classification dataset to make a huge and chaotic dataset
Cifar10 - raw type - data shape is 3072 ( 1024 red value, 1024 blue value, 1024 green value) - transpose: 3,32,32 >> 32,32,3 - convert: RGB >> BGR
Cifar100 - raw type(RGB) - data shape is 3072 ( 1024 red value, 1024 blue value, 1024 green value) - transpose: 3,32,32 >> 32,32,3 - convert: RGB >> BGR
Flower102 - jpg type(BGR) - label starts from 1
CUHK03 - jpg type(BGR) - index of person name is label. - images of each person are divied into 2 sub-group: train 70% and test 30% Examples: # Total samples: 20 Train: 0 ~ 13 Test 14 ~ 19 # Total samples: 20 Train: 0 ~ 13 Test 14 ~ 19 # Total samples: 18 Train: 0 ~ 11 Test 12 ~ 17 # Total samples: 16 Train: 0 ~ 10 Test 11 ~ 15
lfw: - jpg type(BGR) - index of person name is label. - 5786 people, only 1680 people has more than two picutres. - images of each person are divied into 2 sub-group: train 70% and test 30% Examples: # Total samples: 3 Train: 0 ~ 1 Test 2 ~ 2 # Total samples: 2 Train: 0 ~ 0 Test 1 ~ 1 # Total samples: 4 Train: 0 ~ 1 Test 2 ~ 3 # Total samples: 6 Train: 0 ~ 3 Test 4 ~ 5 - first 30 people are excluded.
mnist: - raw type(GRAY) - data shape is 784 - reshape 784 >> 28,28 - convert GRAY >> BGR
stanford: - jpg type(BGR) - label starts from 1
svhn: - raw type(RGB) - data shape is (32,32,3,n) - convert RGB >> BGR
cifar10 ├── batches.meta ├── cifar-10-python.tar.gz ├── data_batch_1 ├── data_batch_2 ├── data_batch_3 ├── data_batch_4 ├── data_batch_5 ├── readme.html └── test_batch
cifar100 ├── cifar-100-python.tar.gz ├── file.txt~ ├── meta ├── test └── train
cuhk03/ ├── cuhk-03.mat ├── cuhk03_release.zip ├── open-reid │?? ├── docs │?? ├── examples │?? ├── images │?? ├── LICENSE │?? ├── meta.json │?? ├── raw │?? ├── README.md │?? ├── reid │?? ├── setup.cfg │?? ├── setup.py │?? ├── splits.json │?? ├── test │?? └── tmp.py └── README.md
flower/ ├── 102flowers.tgz ├── imagelabels.mat ├── jpg └── setid.mat
lfw/ ├── lfw_funneled └── lfw-funneled.tgz
mnist/ ├── t10k-images-idx3-ubyte ├── t10k-labels-idx1-ubyte ├── train-images-idx3-ubyte └── train-labels-idx1-ubyte
stanford_dog/ ├── annotation.tar ├── file_list.mat ├── Images ├── images.tar ├── lists.tar ├── test_list.mat └── train_list.mat
svhn/ ├── test_32x32.mat └── train_32x32.mat