Table 1 DeepPatent2 compared with major sketch-based and natural image datasets. #Categories means the number of object names.
From: DeepPatent2: A Large-Scale Benchmarking Corpus for Technical Drawing Understanding
Dataset | Size (number of images) | #Categories | Captions | Viewpoints |
|---|---|---|---|---|
ImageNet-1k38 | 2M photos | 1,000 | No | No |
ImageNet-21k6 | 14M photos | 21,000 | No | No |
WebVision15 | 2.4M photos | 1,000 | No | No |
Flickr-8K62 | 8K photos | — | Yes | No |
Flickr-30K63 | 30K photos | — | Yes | No |
ImageNet-Sketch13 | 50K drawings | 1,000 | No | No |
TU-Berlin10 | 20K sketches | 250 | No | No |
QuickDraw9 | 50M + sketches | 345 | No | No |
Sketchy3 | 75K sketches, 12K photos | 125 | No | No |
Sheep 10K11 | 10K sheep sketches | 1 | No | No |
Sketch Flickr15K64 | 330 sketches, 15K photos | 33 | No | No |
CLEF-IP 201112 | 10k diagrams | 9 | No | No |
DeepPatent14 | 350K + drawings | — | No | No |
DeepPatent2 | 2M + drawings | 132,890 | Yes | Yes |
