Image Caption Generation Framework using Hybrid Grey Wolf Optimization and Crow Search Algorithm
Saketh Kamatham · Multimedia Research · 2022
The image captioning process is used to generate an image textual description.To generate caption, image captioning uses both natural language processing as well as computer vision.Nevertheless, most image captioning systems present indistinct depictions about objects such as "Man", "woman", "group of people", "building" and so on.Therefore, an intelligent based image captioning technique is developed in this work.The proposed technique consists of some steps such as the formation of the sentence, generation of words, and generation of the caption.At first, the input image is fed to a Deep learning classifier named Convolutional Neural Network (CNN).Here, the main aim of the classifier is used to train the appropriate words which are associated with the image, and can simply classify related words of a given image.Furthermore, the Long-Short Term Memory (LSTM) technique is exploited to form a set of sentences with generated words.Subsequently, the Maximum Like hood (ML) function is exploited to calculate formed sentences likelihood, as well as sentences with maximum probability is exploited that is furthermore exploited to generate a visual illustration of the scene regarding the image caption.The major objective of this work is to improve CNN performance by optimally tuning its activation function and weight.Therefore, for this optimal selection, this work adopts a novel improved optimization approach named Grey Wolf Optimization (GWO) algorithm and Crow Search Algorithm (CSA), which is termed as the Hybrid GWO-CSA algorithm.At last, the adopted captioning technique performance is evaluated with other existing techniques regarding the statistical analysis.