Learning to Group Objects
| Autor: | Victoria Yanulevskaya, Nicu Sebe, Jasper Uijlings |
|---|---|
| Rok vydání: | 2014 |
| Předmět: |
Histograms
Computer science object localisation Selective Search computer vision technique Feature extraction higher quality object hypotheses generation Machine learning computer.software_genre computer vision class-independent object regions random forest training Image color analysis Merging Histogram Radio frequency regions Segmentation image segmentation object grouping computer.programming_language learning business.industry feature extraction region-feature extraction Class independent object proposals segmentation object detection computer vision algorithms region grouping Pattern recognition image parts Image segmentation Pascal (programming language) semantic segmentation Object detection Random forest trees (mathematics) oversegmentation PASCAL dataset learning (artificial intelligence) Artificial intelligence business computer Group object |
| Zdroj: | CVPR Yanulevskaya, V, Uijlings, J R R & Sebe, N 2014, Learning to Group Objects . in Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on . Institute of Electrical and Electronics Engineers (IEEE), pp. 3134-3141 . https://doi.org/10.1109/CVPR.2014.401 |
| DOI: | 10.1109/cvpr.2014.401 |
| Popis: | This paper presents a novel method to generate a hypothesis set of class-independent object regions. It has been shown that such object regions can be used to focus computer vision techniques on the parts of an image that matter most leading to significant improvements in both object localisation and semantic segmentation in recent years. Of course, the higher quality of class-independent object regions, the better subsequent computer vision algorithms can perform. In this paper we focus on generating higher quality object hypotheses. We start from an oversegmentation for which we propose to extract a wide variety of region-features. We group regions together in a hierarchical fashion, for which we train a Random Forest which predicts at each stage of the hierarchy the best possible merge. Hence unlike other approaches, we use relatively powerful features and classifiers at an early stage of the generation of likely object regions. Finally, we identify and combine stable regions in order to capture objects which consist of dissimilar parts. We show on the PASCAL 2007 and 2012 datasets that our method yields higher quality regions than competing approaches while it is at the same time more computationally efficient. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|