cvpr'14 BING objectness

参考代码

https://github.com/torrvision/Objectness

调试到能运行，花了不少时间，中间读数据集报错这些都能出现的，可怕。

编译成功后，直接运行bing_linux，得到输出

 

Dataset:`/home/zherlock/Dataset/CV/VOC2007/' with 2501 training and 4952 testing
WinRecall.m Base = 2, W = 8, NSS = 2, perSz = 130

 
Learning stage I takes 5.27422 seconds...
Learning stage II takes 113.173 seconds...
Start predicting
Average time for predicting an image (MAXBGR) is 0.0426362s
1:0.249,0.304    10:0.478,0.448    100:0.795,0.607    1000:0.96,0.669    2000:0.97,0.673    3000:0.97,0.673    4000:0.97,0.673    5000:0.97,0.673

 

objectness的原文，我是没有通读过的，大概也是一个训练模型的思想，好像是两级的SVM。

emmmmmmmmmmmmm

接下来，