Faculty Publications

Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736

Publications by NITK Faculty

Browse

Filters

2023 - 20242

Settings

Author: search.filters.author.Sowmya Kamath, S.Subject: search.filters.subject.Deep learning

Search Results

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Predicting Air Quality Index with Recurrent Neural Networks and Meta-heuristic Algorithms
    (Institute of Electrical and Electronics Engineers Inc., 2024) Jayanth, P.; Sowmya Kamath, S.
    Millions of people worldwide suffer from the impacts of air pollution, a significant health risk. The metric Air Quality Index (AQI) serves as a crucial tool, providing valuable insights into current air quality conditions and potential health risks. This study utilizes two datasets: one from Wuhan City and the other from Shanghai. The features utilized for forecasting the AQI include PM2.5, PM10, SO2, NO2, O3, CO, l-temp, h-temp, temp, wet, wind, Hecto-pascal Pressure Unit (hpa), visibility, precipitation, and cloud content. This work focuses on developing models to predict AQI for a given data by comparing Long Short Term Memory (LSTM) and its variants, including Bidirectional LSTM (BiLSTM), Stacked LSTM, and Gated Recurrent Unit (GRU) models. Additionally, Particle Swarm Optimization is utilized as an evolutionary feature selection method. © 2024 IEEE.
  • No Thumbnail Available
    Item
    Automated hard exudate segmentation using neural encoders and attention mechanisms for diabetic retinopathy diagnosis
    (Inderscience Publishers, 2023) Gawas, P.; Sowmya Kamath, S.
    Diabetic retinopathy (DR) is a complication caused by increased blood glucose levels, which causes retinal damage in diabetic patients’ eyes. If not discovered and treated early, it can lead to vision loss. Hard exudates (HE) are one of its characteristic signs. Identification of HE is a paramount step in early diagnosis of DR. In this work, the suitability of U-Net-based deep CNN with different encoder configurations and attention gates (AG) is experimented, for HE segmentation. The proposed models were benchmarked on the standard IDRiD dataset. To overcome the challenges related to the limited dataset, data augmentation techniques were also applied to generate image patches and used for model training. Extensive experiments on the dataset revealed that U-Net with AG achieved an accuracy of 98.8%. The U-Net with ResNet50 as the encoder backbone achieved an accuracy of 98.64%. The findings show that the presented models are effective and suitable for early-stage clinical diagnosis. © © 2023 Inderscience Enterprises Ltd.