Salient object based visual sentiment analysis by combining deep features and handcrafted features

With the rapid growth of social networks, the visual sentiment analysis has quickly emerged for opinion mining. Recent study reveals that the sentiments conveyed by some images are related to salient objects in them, we propose a scheme for visual sentiment analysis that combines deep and handcrafte...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2022-03, Vol.81 (6), p.7941-7955
Main Authors: Sowmyayani, S., Rani, P. Arockia Jansi
Format: Article
Language:English
Subjects:
Animal cognition
Animal memory
Artificial neural networks
Computer Communication Networks
Computer Science
Data mining
Data Structures and Information Theory
Datasets
Feature extraction
Multimedia Information Systems
Object recognition
Salience
Sentiment analysis
Social networks
Special Purpose and Application-Based Systems
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the rapid growth of social networks, the visual sentiment analysis has quickly emerged for opinion mining. Recent study reveals that the sentiments conveyed by some images are related to salient objects in them, we propose a scheme for visual sentiment analysis that combines deep and handcrafted features. First, the salient objects are identified from the entire images. Then a pre-trained model such as VGG16 is used to extract deep features from the salient objects. In addition, hand-crafted features such as Visual texture, Colourfulness, Complexity and Fourier Sigma are extracted from all the salient objects. Deep features are combined individually with all the handcrafted features and the performance is measured. The sentiment is predicted using Convolutional Neural Network Classifier. The proposed method is tested on ArtPhoto, Emotion6, Abstract, IAPS datasets, Flickr and Flickr & Instagram datasets. The experimental results substantially proved that the proposed method achieves higher accuracy than other methods.
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-022-11982-5