Boundary-Enhanced Attention for Satellite Imagery dataScience

Satellite image classification presents unique challenges distinct from traditional urban scene datasets, including significant class imbalance and the scarcity of comprehensive examples within single frames. While recent advancements in semantic segmentation and metric learning have shown promise in urban scene datasets, their direct applicability to satellite image classification remains uncertain. This paper introduces a novel approach, the Boundary Attention (BA) Loss , specifically designed to address these challenges in satellite imagery. BA emphasizes the significance of boundary regions within satellite imagery, aiming to mitigate information relation complexity by directing enhanced attention to minority classes and improving attention mechanisms along class boundaries. Through comprehensive experimental evaluation and comparison with existing methods, this paper demonstrates the effectiveness and adaptability of the BA method, paving the way for more accurate and robust satellite image classification systems. The proposed BA method offers a tailored solution that stands to significantly improve classification accuracy in the context of satellite image analysis.

Real-Time Adaptive model for Satellite Image Classification in Dynamic Disaster Environments

During catastrophic events like floods or wildfires, satellite imagery is essential for comprehending the conditions on the ground. Unfortunately, conventional computer vision models find it difficult to adjust to the quickly changing environment that occurs during catastrophes. As a result, the precision with which impacted regions are classified is reduced, making it more difficult to identify flooded areas or destroyed buildings. This research’s main goal is to overcome these obstacles by creating an adaptive computer vision model that is specially made for classifying satellite images. This model will have the capability to generalize across a variety of disaster scenarios and will dynamically adapt to changing conditions by including real-time adaptation mechanisms. As a result, during catastrophes, we will be able to use satellite imagery to gather more precise and fast information, which will increase human safety and disaster response. Our earlier research on the categorization of urban buildup, and land usage and land cover will be expanded upon in this study.

Related publications

  1. Automatic Detection of Natural Disaster Effect on Paddy Field from Satellite Images using Deep Learning Techniques, 8th International Conference on Control and Robotics Engineering (ICCRE), Niigata, Japan, April 21-23, 2023.
  2. Attention Toward Neighbors: A Context-Aware Framework of High-Resolution Image Segmentation, 28th IEEE International Conference on Image Processing (ICIP), Anchorage, Alaska, USA, 2021
  3. Understanding the Urban Environment from Satellite Images with New Classification Method—Focusing on Formality and Informality, Sustainability, vol 14, issue 7, MDPI, 2022
  4. LULC Segmentation of RGB Satellite Image using FCN-8″, in the proceedings of 3rd SLAAI International Conference on Artificial Intelligence, 2019, Sri Lanka and in Communications in Computer and Information Science, Book Chapter, Springer, 2019.


Image Generation Using Variational Autoencoders and Energy Based Models

Machine learning practitioners have long sought generative models that can accurately estimate the underlying data distribution and are able to produce diverse and semantically meaningful image samples. In this project, we take part in a similar quest. Our research directions include (but not limited to) developing improved variational inference techniques, coming up with more efficient methods to train energy based models and better traversing the latent space with Riemannian geometry and manifold assumption.

Bangla Natural Language Processing

In this project, we intend to build a speaker independent Text-to-Speech (TTS) system in Bangla language. To solve the task, we will utilize the SOTA TTS model, Tacotron 2. This model is a combination of two neural network architectures: a modified Tacotron 2 model which is a recurrent sequence-to-sequence model with attention that generates mel-spectrograms from input text. And, a flow-based neural network model named WaveGlow. In this regard, we have created a multi-speaker TTS dataset for Bangladesh Bengali (bn-BD) and Indian Bengali (bn-IN) from Open Speech and Language Resources (OpenSLR) dataset. In our initial experiment, we are interested in the bn-BD dataset. It has audio data of 6 different speakers and corresponding text. Dr. Md Iftekhar Tanveer and Dr. Syeda Sakira Hassan are collaborating with us in this project.

Deep Learning On Graph Data

Nowadays, Graph Representation Learning has seen incredible success across a wide range of modern applications such as recommendation systems, drug discovery, computer vision, particle physics, combinatorial optimization, human activity recognition, etc. In general, a graph consists of nodes and edges where edges depict the relationship between nodes. For instance, the users are nodes in a social network, and edges represent their friendship/interactions. In brief, Graph Neural Networks (GNNs) are designed to compute the low dimensional meaningful embeddings for nodes/graphs through utilizing the structural information and node attributes (if available). These latent representations are used in different downstream tasks e.g., node classification, graph classification, link prediction, community detection/clustering, etc. The high-impact applications of GNNs are leading the research towards developing more advanced deep learning methods. In the high-level view, we aim to leverage probabilistic graphical models, self-supervision learning, contrastive learning, tensor decomposition, and Bayesian modeling into developing more robust, interpretable, and explainable GNNs.

Traffic Forecasting Using Graph Convolutional Network

Intelligent Transportation System (ITS) is being developed in many countries around the world and traffic forecasting lies in the heart of ITS. Traffic intensity is determined by the average speed of vehicles passing through observed road junctions in a traffic network at each time interval and the goal of traffic forecasting is to predict the traffic intensity in near future by observing the traffic data from the past and current time along with the physical traffic network. Research in deep learning models to forecast traffic intensities has gained great attention in recent years due to their capability to capture the complex spatio-temporal relationships within the traffic data. However, the task of traffic forecasting is challenging because there lies a complex spatio-temporal relationship in traffic data as the traffic within a busy city changes heavily in different locations throughout different periods in a day. As the traffic junctions and roads between them can be considered as a graph, we employ graph neural networks to model the traffic forecasting problem which have been published in IJCNN 2021 and PAKDD 2021. Also, to capture the repetitive traffic pattern we consider the traffic data of the past few days instead of considering the traffic data of the past day only. With rigorous experiments we have produced better results for accurate traffic forecasting.