DATASET.md

Set-up Dataset

1. Download dataset.npy into a new directory dataset/.
2. Load dataset using dataset = np.load('dataset/dataset.npy', allow_pickle=True)[()].

Datasets

• CIFAR10
  • 
  • Image shape: (32, 32, 3)
  • No. classes: 9
  • Classes: airplane, car, bird, cat, dog, frog, horse, ship, truck
  • Count per class (train/test):
    • airplane - 5000/1000
    • car - 5000/1000
    • bird - 5000/1000
    • cat - 5000/1000
    • dog - 5000/1000
    • frog - 5000/1000
    • horse - 5000/1000
    • ship - 5000/1000
    • truck - 5000/1000
  • Dataset size (train/test): 45000/9000
  • Download - Run download-cifar.ipynb, then run dataset.ipynb.
  • Website
• QuickDraw
  • 
  • Image shape: (28, 28)
  • No. classes: 9
  • Classes: airplane, bird, car, cat, ship, dog, frog, horse, truck
  • Count per class:
    • airplane - 151623
    • bird - 133572
    • car - 182764
    • cat - 123202
    • ship - 123410
    • dog - 152159
    • frog - 159047
    • horse - 178286
    • truck - 131354
  • Dataset size: 1335417
  • Code
  • Dataset (Numpy 28x28 grayscale bitmap .npy)
• Sketchy
  • 
  • 
  • Real
    • Image shape: (256, 256, 3)
    • No. classes: 8
    • Classes: airplane, car, cat, dog, frog, horse, truck, bird
    • Count per class:
      • airplane - 100
      • car - 100
      • cat - 100
      • dog - 100
      • frog - 100
      • horse - 100
      • truck - 100
      • bird - 100
    • Dataset size: 800
  • Doodle
    • Image shape: (256, 256)
    • No. classes: 8
    • Classes: airplane, car, cat, dog, frog, horse, truck, bird
    • Count per class:
      • airplane - 528
      • car - 534
      • cat - 512
      • dog - 512
      • frog - 502
      • horse - 525
      • truck - 524
      • bird - 504
  • Dataset size: 4141
  • Paper
  • Website
  • Code
  • Dataset (Sketches and Photos)
  • Supplementary Report
• TUBerlin
  • 
  • Image shape: (1111, 1111)
  • No. classes: 8
  • Classes: airplane, car, cat, dog, bird, frog, horse, truck
  • Count per class:
    • airplane - 80
    • car - 80
    • cat - 80
    • dog - 80
    • bird - 80
    • frog - 80
    • horse - 80
    • truck - 80
  • Dataset size: 640
  • Paper
  • Website
  • Dataset (Sketches in png)
• Google Doodle
  • 
  • Image shape: (256, 256)
  • No. classes: 9
  • Classes: airplane, car, bird, cat, dog, frog, horse, ship, truck
  • Count per class:
    • airplane - 500
    • car - 500
    • bird - 500
    • cat - 500
    • dog - 500
    • frog - 500
    • horse - 500
    • ship - 500
    • truck - 500
  • Dataset size: 4500
• Google Real
  • 
  • Image shape: (256, 256, 3)
  • No. classes: 9
  • Classes: airplane, car, bird, cat, dog, frog, horse, ship, truck
  • Count per class:
    • airplane - 100
    • car - 100
    • bird - 100
    • cat - 100
    • dog - 100
    • frog - 100
    • horse - 100
    • ship - 100
    • truck - 100
  • Dataset size: 900