Abstract:In image classification, feature combination is often used to combine the merits of multiple complementary features and improve the classification accuracy compared with one single feature. Existing feature combination algorithms, e.g., multiple kernel learning, usually determine the weights of features based on the optimization with respect to some classifier-dependent objective function. These algorithms are often computationally expensive, and in some cases are found to perform no better than simple baselines. In this paper, we solve the feature combination problem from a totally different perspective. Our algorithm is based on the simple idea of combining only base kernels suitable to be combined. Since the very aim of feature combination is to obtain the highest possible classification accuracy, we measure the combination suitableness of two base kernels by the maximum possible cross-validation accuracy of their combined kernel. By regarding the pairwise suitableness as the kernel adjacency, we obtain a weighted graph of all base kernels and find that the base kernels suitable to be combined correspond to a cluster in the graph. We then use the dominant sets algorithm to find the cluster and determine the weights of base kernels automatically. In this way, we transform the kernel combination problem into a clustering one. Our algorithm can be implemented in parallel easily and the running time can be adjusted based on available memory to a large extent. In experiments on several data sets, our algorithm generates comparable classification accuracy with the state of the art.

Improve Scene Classification by Using Feature and Kernel Combination.

Image Classification Method by Combining Multi-features and Sparse Coding

Fusing Deep Features by Kernel Collaborative Representation for Remote Sensing Scene Classification

Scene Classification Using Low-Level Feature and Intermediate Feature

Scene Classification Using Multi-Resolution Low-Level Feature Combination

Higher-level Feature Combination Via Multiple Kernel Learning for Image Classification

Research of improving semantic image segmentation based on a feature fusion model

A Multi-Instance Multi-Label Scene Classification Method Based on Multi-Kernel Fusion

Feature Combination Via Clustering

Scene Classification Combining Low-level and Semantic Modeling Strategies

Scene classification using a multi-resolution bag-of-features model

Scene classification based on coding layer feature combination

Scene classification with improved AlexNet model

Boosting Classifiers for Scene Category Recognition.

Feature Combination and the kNN Framework in Object Classification

Local, Mid-Level and Convolutional Features Fusion Using Multiple Kernel Learning for Image Classification

Image Classification Based on KPCA and SVM with Randomized Hyper-parameter Optimization

Evaluating Classifier Combination in Object Classification

Learning the Classifier Combination for Image Classification

Scene Classification Using Local and Global Features with Collaborative Representation Fusion

An Effective Image Representation Method Using Kernel Classification