Object Classification using Ensemble of Local and Deep Features

TL;DR

This paper introduces an ensemble approach combining local and deep features for improved object classification, highlighting the complementary nature of features from different CNN layers and architectures.

Contribution

It demonstrates that combining local features with deep CNN features enhances classification performance and that intermediate CNN layers provide valuable information beyond fully connected layers.

Findings

Deep network features complement local features.

Intermediate CNN layers improve classification accuracy.

Features from various CNN architectures encode distinctive information.

Abstract

In this paper we propose an ensemble of local and deep features for object classification. We also compare and contrast effectiveness of feature representation capability of various layers of convolutional neural network. We demonstrate with extensive experiments for object classification that the representation capability of features from deep networks can be complemented with information captured from local features. We also find out that features from various deep convolutional networks encode distinctive characteristic information. We establish that, as opposed to conventional practice, intermediate layers of deep networks can augment the classification capabilities of features obtained from fully connected layers.