Transfer learning for image classification using VGG19: Caltech-101 image data set

Abstract

Image classification is getting more attention in the area of computer vision. During the past few years, a lot of research has been done on image classification using classical machine learning and deep learning techniques. Presently, deep learningbased techniques have given stupendous results. The performance of a classification system depends on the quality of features extracted from an image. The better is the quality of extracted features, the more the accuracy will be. Although, numerous deep learning-based methods have shown enormous performance in image classification, still due to various challenges deep learning methods are not able to extract all the important information from the image. This results in a reduction in overall classification accuracy. The goal of the present research is to improve the image classification performance by combining the deep features extracted using popular deep convolutional neural network, VGG19, and various handcrafted feature extraction methods, i.e., SIFT, SURF, ORB, and Shi-Tomasi corner detector algorithm. Further, the extracted features from these methods are classified using various machine learning classification methods, i.e., Gaussian Naïve Bayes, Decision Tree, Random Forest, and eXtreme Gradient Boosting (XGBClassifier) classifier. The experiment is carried out on a benchmark dataset Caltech-101. The experimental results indicate that Random Forest using the combined features give 93.73% accuracy and outperforms other classifiers and methods proposed by other authors. The paper concludes that a single feature extractor whether shallow or deep is not enough to achieve satisfactory results. So, a combined approach using deep learning features and traditional handcrafted features is better for image classification.

Synthesis of Microscopic Cell Images Obtained from Bone Marrow Aspirate Smears through Generative Adversarial Networks

“…In Table 5 and Table 6 , we have presented the performance of transfer learning models such as InceptionV3 [ 31 ], ResNet [ 32 ], VGG16 [ 33 ], CNN [ 34 ], Xception [ 35 ] and VGG19 [ 36 ] to measure how accurately the classifiers can classify the cell types after training it with original and synthetic data separately. We have used precision and recall as evaluation metrics.…”

Diverse Approach on Image Categorization Using Transfer Learning Methods

“…Authors achieved an accuracy of over 80% using ResNet pretrained networks. Authors of the paper [13] used InceptionV3, a special pre-trained network, to identify pulmonary disease. A number of classifiers were used after separating the features, including SVM, softmax, etc., and achieved an accuracy of around 95.41%.…”

Birdsongs recognition based on ensemble ELM with multi-strategy differential evolution

“…At present, this study only uses MFCC as the feature parameter, and the method proposed in the literature can be tried to extract the features of multiple views for bird audios, and birdsongs recognition can be carried out by combining a variety of different feature parameters. Both deep learning and traditional machine learning are now widely used for object recognition 31 – 33 . Deep learning can also extract low-level features of research objects while classifying.…”

“…Handcrafted features can retain the characteristics of the research object itself, and combine the features extracted by deep learning with handcrafted features to better express the specific information of the object. Literature 31 explored two methods of deep learning and machine learning, and combined traditional features and CNN features, and achieved good results. In the future, we will conduct feature fusion with the representation features extracted by deep learning and traditional hand-extracted features to improve the accuracy of of birdsongs recognition.…”