For object detection, we first need to understand what datasets are commonly used for this task in deep learning. Popular databases for object recognition and object detection are as follow.
| Dataset Name | Total Images | Categories | Images Per Category | Objects Per Image | Image Size | Started Year |
|---|---|---|---|---|---|---|
| MNIST | 60, 000 | 10 | 6, 000 | 1 | 28 × 28 | 1998 |
| Caltech101 | 9, 145 | 101 | 40 ∼ 800 | 1 | 300 × 200 | 2004 |
| Caltech256 | 30, 607 | 256 | 80+ | 1 | 300 × 200 | 2007 |
| Scenes15 | 4, 485 | 15 | 200 ∼ 400 | —— | 256 × 256 | 2006 |
| Tiny Images | 79 millions+ | 53, 464 | —— | —— | 32 × 32 | 2006 |
| PASCAL VOC (2007) | 9, 963 | 20 | 96 ∼ 2008 | 2.5 | 470 × 380 | 2005 |
| PASCAL VOC (2012) | 11, 540 | 20 | 303 ∼ 4087 | 2.4 | 470 × 380 | 2005 |
| SUN | 131, 072 | 908 | —— | 16.8 | 500 × 300 | 2010 |
| ImageNet | 14 millions+ | 21, 841 | —— | 1.5 | 500 × 400 | 2009 |
| KITTI Vision | 14999 | 5 | —— | —— | —— | 2012 |
| Oxford-IIIT Pet | 7, 349 | 37 | 184 ∼ 200 | 1 | —— | 2012 |
| MS COCO | 328, 000+ | 91 | —— | 7.3 | 640 × 480 | 2014 |
| Places | 10 millions+ | 434 | —— | —— | 256 × 256 | 2014 |
| Open Images | 9 millions+ | 6000+ | —— | —— | varied | 2017 |