Generalized Sparselet Models for Real-Time Multiclass Object Recognition

TitleGeneralized Sparselet Models for Real-Time Multiclass Object Recognition
Publication TypeJournal Article
Year of Publication2015
AuthorsSong, H. Oh, Girshick R., Zickler S., Geyer C., Felzenszwalb P., & Darrell T.
Published inIEEE Transactions on Pattern Analysis and Machine Intelligence
Date Published05/2015
KeywordsCaltech101 dataset, Caltech256 dataset, Computational modeling, Deformable models, Deformable Part Models, Dictionaries, frequency 5 Hz, generalized sparselet model, image reconstruction, image representation, ImageNet dataset, laptop computer, laptop computers, multiclass inference, multiconvolutional inference, Object Detection, object recognition, parallel processing, parallelism, PASCAL VOC, real-time multiclass object detection, real-time multiclass object recognition, real-time systems, real-time vision, reconstruction sparsity, shared representation, Sparse Coding, Sparse matrices, standard structured output prediction formulation, Vectors

The problem of real-time multiclass object recognition is of great practical importance in object recognition. In this paper, we describe a framework that simultaneously utilizes shared representation, reconstruction sparsity, and parallelism to enable real-time multiclass object detection with deformable part models at 5Hz on a laptop computer with almost no decrease in task performance. Our framework is trained in the standard structured output prediction formulation and is generically applicable for speeding up object recognition systems where the computational bottleneck is in multiclass, multi-convolutional inference. We experimentally demonstrate the efficiency and task performance of our method on PASCAL VOC, subset of ImageNet, Caltech101 and Caltech256 dataset.

ICSI Research Group