Improved Image Captioning Using Associative Correlation
M.S. Minu1, Akilan Ganesan2, Ram Abhilash V3, M Mageswaran4
1Akialn Ganesan, Department of Computer science, SRM institute of science and technology, Chennai (Tamil Nadu), India.
2Ram Abhilash, Department of Computer science, SRM institute of science and technology, Chennai (Tamil Nadu), India.
3M Mageswaran, Department of Computer science, SRM institute of science and technology, Chennai (Tamil Nadu), India.
Manuscript received on 18 April 2019 | Revised Manuscript received on 25 April 2019 | Manuscript published on 30 April 2019 | PP: 1283-1288 | Volume-8 Issue-4, April 2019 | Retrieval Number: D6607048419/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The production of appropriate captions to a given image is the characteristic feature of Image captioning. It helps to identify the notable areas of a certain image. Even Though neural networks do have, as of late accomplished promising outcomes, a key issue that still exist is that they can just portray ideas found in trained image-sentence data sets. Proficient learning and portrayal of novel ideas has accordingly been the centre of focus of recent researches. This would help to ease the costly labour of naming data and their sets. The authors propose a focused search by using the data gained from the individual areas of the image with different sources of information from various data sets to train the existing model to come up with captions that are outside the image captioning datasets. Our model utilizes semantic data to produce subtitles for several item classes in ImageNet object identification dataset. Both programmed assessments and human assessments demonstrate that our model significantly outflanks earlier work in having the capacity to portray a lot more Classifications of articles which are outside the scope of the existing datasets.
Keywords: Image Captioning, Novel Concept, Visual Attention.
Scope of the Article: Classifications