《Learning Deep Features for Scene Recognition using Places Database》记录

《Learning Deep Features for Scene Recognition using Places Database》记录

***scene dataset:

scene15: 15 categories, a few hundred images per class, 95% accuracy.

MIT indoor67:  67 categories, indoor scene

SUM db: 397 categories, 100 images per class

places db: 700w, 476 categories.

生成方式：

1. 生成section，来源于sum db。

2. amt，进行2轮label，(1) default no, choose yes. (2) from labeled yes image, default yes choose no.   90% accuracy.   750 image  + 30 positive samples + 30 negative samples

***evalue db

a good database should be dense (with a high degree of data concentration), and diverse (it should include a high variability of appearances and viewpoints)