MR WHY

I am
Wang Hongyang

Wechat Official Account
Find fun things here!

Contour Box Rough Code Release - Contour Box 粗糙代码

Related Paper

Contour Box: Rejecting Object Proposals Without Explicit Closed Contours, by Cewu Lu et al.

Abstract and Main Idea

Abstract: Closed contour is an important objectness indicator. We propose a new measure subject to the completeness and tightness constraints, where the optimized closed contour should be tightly bounded within an object proposal. The closed contour measure is defined using closed path inte- gral, and we solve the optimization problem efficiently in polar coordinate system with a global optimum guaran- teed. Extensive experiments show that our method can re- ject a large number of false proposals, and achieve over 6% improvement in object recall at the challenging overlap threshold 0.8 on the VOC 2007 test dataset.

Main Idea:

For more detailed introduction, please refer to a previous blog article.

Code

My realized result of Contour Box:


Fig2. Line changes as the parameter of completeness constraint changes. The right one is the best with a relatively strict completeness constraint.

Download Code on Github

Note that this code cannot reach the outcome in the paper on the whole PASCAL VOC2007 dataset. (Although I have tried my best and totally realized according to the paper...)

So the purpose of releasing this code is to save friends time of rewriting the basic codes. And if any friend can get the result in the paper, you can directly pull a requist to github or commit below or contact me via email: h.y.wang@outlook.com

THANK YOU ~ !


Author:Hongyang Wang, http://mr-why.com
Please feel free to repost if obey Creative Commons BY-NC-ND 3.0.

53

分享本文:

TOC