Sparse distributed localized gradient fused features of objects

作者：

Highlights：

• We develop a method for sparse distributed selection and encoding of object features.

• We demonstrate improved object recognition accuracies for ALOI, COIL-100 and PASCAL databases.

• Modular and hierarchical processing of sparse features reported to be advantageous.

• Increased natural variability results in reduced recognition performance.

摘要

Highlights•We develop a method for sparse distributed selection and encoding of object features.•We demonstrate improved object recognition accuracies for ALOI, COIL-100 and PASCAL databases.•Modular and hierarchical processing of sparse features reported to be advantageous.•Increased natural variability results in reduced recognition performance.

论文关键词：Brain-inspired system,Feature fusion,Hierarchy,Modularity,Object feature,Object recognition,Sparse feature

论文评审过程：Received 28 September 2013, Revised 28 September 2014, Accepted 1 October 2014, Available online 31 October 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.10.002