Dhruv Shindhe, S and Faheema, AGJ (2023) Inquisition of Vision Transformer for Content Based Satellite Image Retrieval. In: International Conference on Computer, Communication, and Signal Processing, ICCCSP 2023, pp. 171-182.
Full text not available from this repository.Abstract
Content based satellite image retrieval is among the most vital technologies for military application. Huge volumes of satellite imagery get generated day by day. Management of huge archives of satellite image requires intelligent image retrieval system. In this paper, we have carried out study using SOTA Vision Transformer and other deep CNN based features in conjunction with various matching and indexing methods. We conducted exhaustive experiments by employing various deep features and different variants of Vision Transformer with variety of feature vector matching techniques. Experimental results conducted on various benchmark satellite image retrieval dataset demonstrate the promising capability of the Vision Transformer in conjunction with Ball tree. ResNet & DenseNet have outperformed on two benchmark datasets namely RS19, & PatterNet. ViT has outperformed on RSSC7N dataset & has performed on par with ResNet & DenseNet on UCM dataset. © IFIP International Federation for Information Processing 2023.
| Item Type: | Conference Paper |
|---|---|
| Publication: | IFIP Advances in Information and Communication Technology |
| Publisher: | Springer Science and Business Media Deutschland GmbH |
| Additional Information: | The copyright belongs to the IFIP International Federation for Information Processing. |
| Keywords: | Image retrieval; Military applications; Military photography; Search engines, Ball tree; BERT; Content-based; Densenet; Image retrieval systems; Intelligent image retrieval; Resnet; Satellite image retrieval; Satellite images; ViT, Satellite imagery |
| Department/Centre: | Others |
| Date Deposited: | 15 Jul 2024 12:16 |
| Last Modified: | 15 Jul 2024 12:16 |
| URI: | http://eprints.iisc.ac.in/id/eprint/84368 |
Actions (login required)
 |
View Item |
