Segmentation as selective search for object recognition

Abstract

For object recognition, the current state-of-the-art is based on exhaustive search. However, to enable the use of more expensive features and classifiers and thereby progress beyond the state-of-the-art, a selective search strategy is needed. Therefore, we adapt segmentation as a selective search by reconsidering segmentation: We pro pose to generate many approximate locations over few and precise object delineations because (1) an object whose lo cation is never generated can not be recognised and (2) ap pearance and immediate nearby context are most effective for object recognition. Our method is class-independent and is shown to cover 96.7% of all objects in the Pascal VOC 2007 test set using only 1,536 locations per image. Our selective search enables the use of the more expensive bag-of-words method which we use to substantially improve the state-of-the-art by up to 8.5% for 8 out of 20 classes on the Pascal VOC 2010 detection challenge.

Data Decomposition and Spatial Mixture Modeling for Part Based Model

“…We run the detector implemented by ourself on PASCAL VOC2010. SegAs is the latest result from [19] which mainly focuses on selecting windows with high "objectness" via segmentation. Their object appearance model is based on bag-of-words.…”

“…rescoring and selective window search (e.g., [19]). MKL method with four different features also provides very competitive results, and our system gets better results by nearly 4%.…”

“…DDNoSSM also obtains the second best mean AP. SegAs [19] focuses on selective window sampling via segmentation. The proposed method exceeds [19] by 1.6% without any selective search.…”

“…They propose a coarse-to-fine search strategy to speedup the complex hierarchical part based model. In [19], they propose a selective window sampling strategy via segmentation which makes using multiple features based bag-of-words possible. Those methods indeed reduce search space and speedup training and testing, but they all do not improve models' discrimination.…”

CollageParsing: Nonparametric Scene Parsing by Adaptive Overlapping Windows

“…Other algorithms for generating class-generic object window predictions, such as van de Sande's hierarchical segmentation based windows [30], can also be used in place of objectness at this stage in the pipeline.…”