Soumya Ghosal

Created a self-attention based semantic segmentation method, the proposed method outperformed all the present state of the art methods and achieved record benchmark accuracy for ISPRS Vaihingen & ISPRS Potsdam dataset (91.36 % and 92.41% respectively). The github link is given below. https://github.com/soumyaghosal03/Attention-based-semantic-segmentation-for-Remote-sensing-images

Gathering, processing, storing and making some initial insights about fashion data from social network platforms and websites. The github link is given below. https://github.com/soumyaghosal03/Fashion-Dataset-gathering-filtering-and-storing-in-a-database

1.Worked in projects related to tuning, optimization and designing control system models of prosthetic organs using stochastic and evolutionary algorithms. 2.Published 5 international conference and journal papers and 1 Patent.

Creation of a Content Based Image Retrieval model using Artificial Neural Networks and Fuzzy C Means clustering

Courses taken - Photogrammetric Computer Vision; Semantic 3D/4D City Models; Advanced Methods for Geospatial Analysis; Selected Sections of Geoinformatics – Deep Learning with Convolutional Neural Networks; Computer Graphics 2; Statistic Test Procedure and Analysis of Stochastic Processes; Adjustment Calculation 1 and 2 – Linear and Non-linear adjustment using Gauss- Markov and Gauss- Helmert model; Microwave and Radar Remote Sensing; Computer Vision Seminar; Planetary Geodesy.

Courses taken - Programming in C; Object Oriented Programming; Data Structure and Algorithms; Database Management Systems; Formal Language and Automata theory; Advanced Computer Architecture; Graph theory; Control Systems; Artificial Neural Networks; Web Technology; Computer Networking; Software Engineering; Business Information Systems; Linear Algebra; etc.