0%

计算机视觉--常用数据库

summary

dataset im_size class*num download task example code pretrained_model state of art 实例图片 备注
mnist 28*28 70000
CIFAR-10 32x32x3 10*6000 d Alex,Hinton发布,超小图片
CIFAR-100 32x32 170M 100*600
Pascal VOC (05-12) 2GB voc2012
coco 40GB
imagenet2012 尺寸不固定,但多数比较清晰 1000类,训练集1.2m,验证集50k,测试集100k 层级标签。 评价指标,top5的label包含正确label就算正确 AlexNet
imagenet2016 1400多万幅图片,涵盖2万多个类别 1TB
places
12306 约80*80 12306图片比cifar数据库大多了

分类结果汇总

task_code:

  1. 图像分类(image classification)
  2. 目标检测(object detection)
  3. 目标识别(object recognition)
  4. 语义分割(semantic segmentation)
  5. 实例分割(instance segmentation)

来自pytorch vision的data

  • LSUN http://lsun.cs.princeton.edu`_ dataset
  • Local Image Descriptors Data http://phototour.cs.washington.edu/patches/default.htm`_ Dataset.
  • SEMEION http://archive.ics.uci.edu/ml/datasets/semeion+handwritten+digit`_ Dataset.
  • STL10 https://cs.stanford.edu/~acoates/stl10/`_ Dataset.
  • SVHN http://ufldl.stanford.edu/housenumbers/`_ Dataset.

来自tensorflow的data