Unconstrained Salient Object Detection

We aim to predict bounding box localizations for salient object in unconstrained images. We propose a system that can output a highly reduced set of output windows based on a CNN proposal generation model and a novel proposal subset optimization formulation. Our system significantly outperforms existing methods in localizing dominant objects.



Jianming Zhang, Stan Sclaroff, Zhe Lin, Xiaohui Shen, Brian Price and Radomír Měch. "Unconstrained Salient Object Detection via Proposal Subset Optimization" To appear in Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.




  title={Unconstrained Salient Object Detection via Proposal Subset Optimization},

  author={Zhang, Jianming and Sclaroff, Stan and Lin, Zhe and Shen, Xiaohui and Price, Brian and M\u{e}ch, Radom\'{i}r},

  booktitle={IEEE Conference on Computer Vision and Pattern Recognition(CVPR)},



Contact: jmzhang AT bu.edu


We provide a Matlab implementation of our full system together with pre-trained  CNN models for proposal generation.

Code & CNN models: [Github] (The models will be automatically downloaded by the Matlab scripts.)

CNN models only: [GoogleNet] [VGG16]

Annotation Data

We provide bounding box annotations for the training split of the MSO dataset.

Annotation: [download]