RAHUL S HOSKERI 👋

Objective: Build a query system to return the bird sound spectogram images similar to the query spectogram image.

• Employed the digital image proceesing techniques of spectogram and region of interest calculation on audio files for preprocessing & Gabor texture feature extraction.
• The query system achieved an average accuracy of ≈0.65.

Objective: Train a neural network to parallelize and reduce the calculation time of Solid Oxide Fuel Cell simulation.

• Training a multi layer perceptron model for the simulation calculation achieved an average accuracy of ≈78% and simulation time optimised by ≈ 98%.

Objective: Recognize and translate between the dialects of Dravidian dialect using computer vision/image processing to restore and digitize historic scriptures.

• The old scriptures in Tulu needs to be preserved by translating to the prevailant language to protect the literature hritage of the region.
• Calculating structural similarity using cross-correlation technique achieved an accuracy of ≈ 82% for recognition and translation.

Objective: The classical problem of operational research/approximation algorithms to minimise the distance between the facility and the places served by it.

• Local search (1-Swap) K-median heuristics & Linear programming were used to optimise the solution for 500 cities coordinates in USA.

Objective: Perform text analysis to understand the kind of relationship they share and build a word cloud image to understand the same.

• Text analysisconverstation can help understand the relationship of persons under conversation to help analyse the mood and behaviours.