基于ICA-HMAX的视觉与语义相似性计算模型

Visual and Semantic Similarity Model Based on ICA-HMAX

该文首先简单介绍多层次卷积ICA-HMAX模型;然后提出卷积核的可视化方法实现对不同层次卷积核的理解。提出采用ICA-HMAX模型的S3层特征实现不同类别的视觉的相似度计算。通过在Caltech101数据集上实现对不同的类别之间的视觉相似度计算,并与类别之间的语义相似度进行对比分析进而探讨语义相似与视觉相似之间的关系。

This paper firstly introduces a multi-level convolutional ICA-HMAX model; then propose visualization methods of the convolution kernel convolution kernel to achieve different levels of understanding. Proposed using S3 layer feature of the ICAHMAX model to achieve visual similarity calculation of different categories. By implementing different categories of visual similarity between calculated on Caltech 101 datasets, and compared with the semantic similarity between categories and further explore the relationship between semantic similarity and visual similarity between.