Meghna Kapoor

I am a Ph.D. Research Scholar under the supervision of Dr. Badri N Subudhi and Dr. Ankur Bansal in the EE Department at Indian Institute of Technology Jammu, India since 2021. I primarily work in the area of image processing, computer vision, graph learning, and underwater surveillance.

I have completed my M.E. in Electronics and Communication from Thapar Institute of Engineering and Technology Patiala, India.

Email  /  Google Scholar  /  Linkedin  /  CV

16/06/2025 - One paper accepted in NCVPRIPG 2025. Congratulations Gokul!

07/06/2025 - One paper accepted in IEEE Transactions on Artficial Intelligence. Congratulations Ashutosh and Dakshi

14/05/2025 - Received ACM IARCS and ANRF ITS travel grants to attend ICME 2025

14/04/2025 - Our workshop proposal in underwater advances is accepted in NCVPRIPG 2025 Link

14/03/2025 - One paper accepted in the IEEE International Conference on Multimedia and Expo (ICME) 2025

06/02/2025 - We have filed one US patent with Application number 18230642 Patent Link

04/02/2025 - Received IEEE SPS Travel Grant to attend ICASSP 2025

14/01/2025 - One paper accepted in the Journal of Pattern Recognition (PR).

09/01/2025 - One paper accepted in the Journal of Computer Vision and Image Understanding (CVIU). Congratulations Wieke Prummel !

21/12/2024 - One paper accepted in the International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2025

17/08/2024 - One paper accepted in the International Conference on Pattern Recognition (ICPR) 2024.

06/06/2024 - One paper accepted in Ocean Engineering

20/08/2023 - Received €1000 travel grant for ICCV 2023

07/08/2023 - One paper accepted in IEEE/CVF International conference on computer vision workshops (ICCVW) 2023

02/05/2023 - Received 1400 USD Travel award for CVPR 2023

06/04/2023 - One paper accepted in IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2023

08/12/2021 - Awarded Prime Minister's fellowship scheme for doctoral research

15/01/2021 - Started PhD in the domain of underwater surveillance

I am currently engaged in the field of underwater surveillance, focusing on mitigating degradations in complex underwater environments. My work involves implementing various methodologies to enhance image clarity, preserve details, and accurately detect objects in challenging underwater conditions.

We propose a novel clustering approach based on the assumption that high-quality images contain richer high-level information, which is extracted using a pre-trained model.

We introduce an unsupervised single-image training-based spatially curated perceptual loss (USCPL) model to enhance degraded underwater images.

We propose a semi-supervised graph-learning approach (GraphMOS-U) to segment moving objects in underwater environments.

The proposed approach uses a convolutional backbone to project the image into latent space, where an unsupervised initial graph is constructed. The hypergraph convolution is then utilized to optimize message passing between graph nodes, enhancing the representation of complex relationships of latent space. This helps in the retention of intricate details by modelling two or more latent variables as hyperedge by sharing the information among themselves. Finally, an image generation module maps the enhanced graph representation back to image space.

This study presents an end-to-end moving object detection architecture to analyze intricate underwater scenes. We adhered to a ResNet-50 backbone in the proposed architecture to project the video frame to feature space. Graph learning is used to retain the structural information of the object by projecting from feature space to graph space. Multiple aggregators facilitate the seamless transfer of information among neighbouring nodes, alleviating noise induced by deep architectures. The refactored latent vector is transformed to image space to detect the moving object(s) from the given scene.

The objective of this article is to provide a detailed review of the state-of-the-art underwater surveillance process and the new trends of the same. Underwater surveillance has recently gotten lots of attention because of its potential applications including the security of the coastal border, effective fish farming, deep-sea exploration, preservation of rare aquatic animals, etc.

This paper proposes an encoder-decoder network that preserves the image content, texture, and style while maintaining overall global similarity by capturing the inherent distribution of the training samples. To overcome the deviation due to a change in water type, a classifier network is induced in the latent space of encoder-decoder architecture. The classifier loss and adversarial loss in the classifier network ensure the learning across domains and avoid setting priors on captured distribution. Hence, the proposed model is robust against the change of water type and can be deployed in real-life without re-training.

We propose a unique architecture for underwater object detection where U-Net architecture is considered with the ResNet-50 backbone. Further, the latent space features from the encoder are fed to the decoder through a GraphSage model. GraphSage-based model is explored to reweight the node relationship in non-euclidean space using different aggregator functions and hence characterize the spatio-contextual bonding among the pixels. Further, we explored the dependency on different aggregator functions: mean, max, and LSTM, to evaluate the model's performance.

Thanks for Visiting my Page.