Fruit classification using deep feature maps in the presence of deceptive similar classes

TL;DR

This paper proposes an ensemble deep learning approach that leverages multi-layer CNN features and decision trees to improve classification of deceptively similar fruit classes, outperforming conventional methods.

Contribution

It introduces a novel ensemble method using multi-layer CNN activations and random forests for better discrimination of similar classes.

Findings

Outperforms traditional deep learning models on Fruits-360 dataset.

Utilizes multi-layer CNN features for improved classification accuracy.

Effective in distinguishing deceptive similar classes.

Abstract

Autonomous detection and classification of objects are admired area of research in many industrial applications. Though, humans can distinguish objects with high multi-granular similarities very easily; but for the machines, it is a very challenging task. The convolution neural networks (CNN) have illustrated efficient performance in multi-level representations of objects for classification. Conventionally, the existing deep learning models utilize the transformed features generated by the rearmost layer for training and testing. However, it is evident that this does not work well with multi-granular data, especially, in presence of deceptive similar classes (almost similar but different classes). The objective of the present research is to address the challenge of classification of deceptively similar multi-granular objects with an ensemble approach thfat utilizes activations from multiple layers of CNN (deep features). These multi-layer activations are further utilized to build multiple deep decision trees (known as Random forest) for classification of objects with similar appearance. The Fruits-360 dataset is utilized for evaluation of the proposed approach. With extensive trials it was observed that the proposed model outperformed over the conventional deep learning approaches.