JTJagrati Talreja

Jagrati Talreja

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Jagrati Talreja has a total of 4 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    D-PolTransGAN: Scattering Guided Hybrid Transformer-GAN for SAR-Optical Image Translation Using Dual Polarization SAR

    by Jagrati Talreja, Tewodros Syum Gebre, Leila Hashemi-Beni

    Synthetic Aperture Radar (SAR) provides reliable imaging in all weather and lighting conditions, but its speckled appearance makes visual interpretation difficult and limits its integration with optical-based analyses. To address this challenge, we propose a new SAR-to-optical image translation framework tailored for flood assessment. The proposed model processes SAR VV and VH polarizations through separate feature extraction streams, enabling each stem to learn its unique scattering characteristics. These features are then fused using cross-attention to capture complementary information from both polarizations. The architecture combines convolutional networks for local detail extraction with transformer modules to model global context, all within a generative adversarial learning framework. We train and evaluate the model using the DEEPFLOOD, SEN1FLOODS11 and SEN12MS datasets, which provides well-aligned Sentinel-1 (VV, VH) and Sentinel-2 image pairs. Experimental results show that the proposed method outperforms existing approaches in terms of PSNR, SSIM, and perceptual quality (LPIPS), while preserving fine spatial details critical for flood analysis. Qualitative results on real flood events further demonstrate that the generated optical-like images recover visual structures and contextual information that are difficult to interpret in raw SAR imagery. https://github.com/JagratiTalreja01/D-PolTransGAN


  • Top Cited Papers
    Open Access

    XTNSR: Xception-based transformer network for single image super resolution

    by Jagrati Talreja, Supavadee Aramvith, Takao Onoye

    Single image super resolution has significantly advanced by utilizing transformers-based deep learning algorithms. However, challenges still need to be addressed in handling grid-like image patches with higher computational demands and addressing issues like over-smoothing in visual patches. This paper presents a Deep Learning model for single-image super-resolution. In this paper, we present the XTNSR model, a novel multi-path network architecture that combines Local feature window transformers (LWFT) with Xception blocks for single-image super-resolution. The model processes grid-like image patches effectively and reduces computational complexity by integrating a Patch Embedding layer. Whereas the Xception blocks use depth-wise separable convolutions for hierarchical feature extraction, the LWFT blocks capture long-range dependencies and fine-grained qualities. A multi-layer feature fusion block with skip connections, part of this hybrid architecture, guarantees efficient local and global feature fusion. The experimental results show better performance in Peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and visual quality than the state-of-the-art techniques. By optimizing parameters, the suggested architecture also lowers computational complexity. Overall, the architecture presents a promising approach for advancing image super-resolution capabilities.


  • Open Access

    DHTCUN: Deep Hybrid Transformer CNN U Network for Single-Image Super-Resolution

    by Jagrati Talreja, Supavadee Aramvith, Takao Onoye

    Recent advances in image super-resolution have investigated various transformer and CNN techniques to improve quantitative and perceptual outcomes. Reconstructing high-resolution images from their low-resolution equivalents by combining the power of transformers and CNN has been a crucial task in recent times.We propose a novel U-shaped architecture that integrates transformers and convolutional neural networks (CNNs) to leverage the strengths of both approaches. The network incorporates a novel Parallel Hybrid Transformer CNN Block (PHTCB) on the backbone of the U-shaped design, ensuring computational efficiency and robust hierarchical feature representation. Our architecture incorporates triple-enhanced spatial-attention mechanisms and a Transformer CNN (TCN) Block in PHTCB. The TCN Block helps preserve sharp edges and intricate details often lost in traditional SISR methods and enhances the visual fidelity of the reconstructed high-resolution images. Additionally, we introduce the triple-enhanced spatial attention (TESA) approach that helps precisely localize of important features. Blurring can be reduced for crucial features by focusing on these critical areas because of the network’s ability to control features at various scales. Experiments demonstrate that our proposed method yields better quantitative measurements, including visually appealing high-resolution image reconstructions, peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM).


  • Open Access

    DANS: Deep Attention Network for Single Image Super-Resolution

    by Jagrati Talreja, Supavadee Aramvith, Takao Onoye

    The current advancements in image super-resolution have explored different attention mechanisms to achieve better quantitative and perceptual results. The critical challenge recently is to utilize the potential of attention mechanisms to reconstruct high-resolution images from their low-resolution counterparts. This research proposes a novel method that combines inception blocks, non-local sparse attention, and a U-Net network architecture. The network incorporates the non-local sparse attention on the backbone of symmetric encoder-decoder U-Net structure, which helps to identify long-range dependencies and exploits contextual information while preserving global context. By incorporating skip connections, the network can leverage features at different scales, enhancing the reconstruction of high-frequency information. Additionally, we introduce inception blocks allowing the model to capture information at various levels of abstraction to enhance multi-scale representation learning further. Experimental findings show that our suggested approach produces superior quantitative measurements, such as peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), visual information fidelity (VIF), and visually appealing high-resolution image reconstructions.