Robotics and Computer Vision Laboratory Login  
  Robotics and Computer Vision Laboratory kaist logo
Archive Courses

Home  >  Research  >  Publications

[International Conference] AttentionNet: Aggregating Weak Directions for Accurate Object Detection
Donggeun Yoo , Sunggyun Park , Joon-Young Lee , Anthony Paek , In So Kweon
IEEE International Conference on Computer Vision (ICCV) , December 2015
  1943.pdf 1943.pdf (1.8M) [145]
We present a new detection method using a convolutional neural network (CNN), named AttentionNet. We cast an object detection problem as an iterative classification problem, which is the most suitable form of a CNN. AttentionNet provides quantized weak directions pointing a target object and iterative predictions from AttentionNet converge to an accurate object boundary box. Since AttentionNet is an unified network for object detection, it detects objects without any separated models from initial object proposal to accurate object localization. We evaluate AttentionNet by a human detection task and achieve the state-of-the-art performance of 65% (AP) on PASCAL VOC 2007/2012 with an 8-layered architecture.


Robotics and Computer Vision Laboratory
KAIST | Electrical Engineering | Contact Us | Sitemap