arXiv subjects

Omkar Oak

Publications and source records attributed to Omkar Oak.

4 recordsLinked to original sources

arXiv2026

TERRA-CD: Multi-Temporal Framework for Multi-class and Semantic Change Detection

Urban vegetation monitoring plays a vital role in understanding environmental changes, yet comprehensive datasets for this purpose remain limited. To address this gap, we present the Temporal Remote-sensing Repository for Analyzing Change Detection (TERRA-CD), a benchmark dataset comprising 5,221 Sentinel-2 image pairs from 2019 and 2024, covering 232 cities across the USA and Europe. The dataset features three distinct annotation schemes: 4-class land cover mapping masks, 3-class vegetation change masks, and 13-class semantic change masks capturing all possible land cover transitions. Using various deep learning approaches including Siamese networks, STANet variants, Bi-SRNet, Changemask, Post-Classification Comparison, and HRSCD strategies, we evaluated the dataset's effectiveness for both vegetation Multi-class Change Detection as well as Semantic Change Detection. The proposed dataset and methods are available at https://github.com/omkarsoak/TERRA-CD.

cs.CV↗

arXiv2024

A Temporal Convolutional Network-based Approach for Network Intrusion Detection

Network intrusion detection is critical for securing modern networks, yet the complexity of network traffic poses significant challenges to traditional methods. This study proposes a Temporal Convolutional Network(TCN) model featuring a residual block architecture with dilated convolutions to capture dependencies in network traffic data while ensuring training stability. The TCN's ability to process sequences in parallel enables faster, more accurate sequence modeling than Recurrent Neural Networks. Evaluated on the Edge-IIoTset dataset, which includes 15 classes with normal traffic and 14 cyberattack types, the proposed model achieved an accuracy of 96.72% and a loss of 0.0688, outperforming 1D CNN, CNN-LSTM, CNN-GRU, CNN-BiLSTM, and CNN-GRU-LSTM models. A class-wise classification report, encompassing metrics such as recall, precision, accuracy, and F1-score, demonstrated the TCN model's superior performance across varied attack categories, including Malware, Injection, and DDoS. These results underscore the model's potential in addressing the complexities of network intrusion detection effectively.

cs.CR↗

arXiv2024

A Novel Multivariate Bi-LSTM model for Short-Term Equity Price Forecasting

Prediction models are crucial in the stock market as they aid in forecasting future prices and trends, enabling investors to make informed decisions and manage risks more effectively. In the Indian stock market, where volatility is often high, accurate predictions can provide a significant edge in capitalizing on market movements. While various models like regression and Artificial Neural Networks (ANNs) have been explored for this purpose, studies have shown that Long Short-Term Memory networks (LSTMs) are the most effective. This is because they can capture complex temporal dependencies present in financial data. This paper presents a Bidirectional Multivariate LSTM model designed to predict short-term stock prices of Indian companies in the NIFTY 100 across four major sectors. Both Univariate LSTM and Univariate Bidirectional LSTM models were evaluated based on R2 score, RMSE, MSE, MAE, and MAPE. To improve predictive accuracy, the analysis was extended to multivariate data. Additionally, 12 technical indicators, having high correlation values with the close price(greater than 0.99) including EMA5, SMA5, TRIMA5, KAMA10 and the Bollinger Bands were selected as variables to further optimize the prediction models. The proposed Bidirectional Multivariate LSTM model, when applied to a dataset containing these indicators, achieved an exceptionally high average R2 score of 99.4779% across the four stocks, which is 3.9833% higher than that of the Unidirectional Multivariate LSTM without technical indicators. The proposed model has an average RMSE of 0.0103955, an average MAE of 0.007485 and an average MAPE of 1.1635%. This highlights the model's exceptional forecasting accuracy and emphasizes its potential to improve short-term trading strategies.

cs.CE↗

arXiv2024

A Comparative Analysis of CNN-based Deep Learning Models for Landslide Detection

Landslides inflict substantial societal and economic damage, underscoring their global significance as recurrent and destructive natural disasters. Recent landslides in northern parts of India and Nepal have caused significant disruption, damaging infrastructure and posing threats to local communities. Convolutional Neural Networks (CNNs), a type of deep learning technique, have shown remarkable success in image processing. Because of their sophisticated architectures, advanced CNN-based models perform better in landslide detection than conventional algorithms. The purpose of this work is to investigate CNNs' potential in more detail, with an emphasis on comparison of CNN based models for better landslide detection. We compared four traditional semantic segmentation models (U-Net, LinkNet, PSPNet, and FPN) and utilized the ResNet50 backbone encoder to implement them. Moreover, we have experimented with the hyperparameters such as learning rates, batch sizes, and regularization techniques to fine-tune the models. We have computed the confusion matrix for each model and used performance metrics including precision, recall and f1-score to evaluate and compare the deep learning models. According to the experimental results, LinkNet gave the best results among the four models having an Accuracy of 97.49% and a F1-score of 85.7% (with 84.49% precision, 87.07% recall). We have also presented a comprehensive comparison of all pixel-wise confusion matrix results and the time taken to train each model.

cs.CV↗