深度学习——图像分类与检测领域常用数据集汇总
李是李雅普诺夫的李 人气:2图像分类与检测领域常用数据集汇总
| 数据库 | 图像数目 | 类别数目 | 每类图像数目 | 图像大小（pixel） | 数据库论文（点击即可获取论文，如果无法下载，推荐使用SCI-HUB） |
| MNIST | 60000 | 10 | 6000 |
28x28 |
Gradient-based learning applied to document recognition |
| CIFAR-10 | 60000 | 10 | 6000 | 32x32 | Learning multiple layers of features from tiny images |
| MPEG-7 | 1400 | 70 | 20 | 256x256-650x600 | Shape descriptors for non-rigid shapes with a single closed contour |
| 15 Scenes | 4485 | 15 | 200-400 | 约 300x250 | Beyond bags of features: Spatial pyramid matching for recognizing natural scene categories |
| Caltech-101 | 9146 | 101 | 40-800 | 约 300x200 | Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories |
| Caltech-256 | 30607 | 256 | >80 | 约 300x200 | Caltech-256 object category dataset |
| PASCAL VOC 2007 | 9963 | 20 | 96-2008 | 约 470x380 |
The PASCAL Visual Object Classes (VOC) Challenge |
| SUN397 | 108754 | 397 | >100 | 约 500x300 | Sun database: Large-scale scene recognition from abbey to zoo |
| SUN2012 | 16873 | 8 | 2000 | 约 500x300 | Sun database: Large-scale scene recognition from abbey to zoo |
| Tiny Images | 7900万 | 75062 | - | 32x32 | 80 million tiny images: A large data set for nonparametric object and scene recognition |
| ImageNet-1000 | 120万 | 1000 | - | 约 500x400 | Imagenet: A large-scale hierarchical image database |
| ImageNet | 1400万 | 10万 | 1000 | 约 500x400 | Imagenet: A large-scale hierarchical image database |
参考文献：黄凯奇, 任伟强, 谭铁牛. 图像物体分类与检测算法综述[J]. 计算机学报, 2014, 37(6): 1225-1240.
加载全部内容