Synergetic Multi-Semantic Multi-Instance Learning for Scene Recognition
Wen Gang Feng, Xue Chen · Applied Mechanics and Materials · 2012
In this paper, the problem of scene representation is modeled by simultaneously considering the stimulus-driven and instance-related factors in a probabilistic framework. In this framework, a stimulus-driven component simulates the low-level processes in human vision system using semantic constrain; while a instance-related component simulate the high-level processes to bias the competition of the input features. We interpret the synergetic multi-semantic multi-instance learning on five scene database of LabelMe benchmark, and validate scene classification on the fifteen scene database via the SVM inference with comparison to the state-of-arts methods.