π-SUB: A Physics Informed Synthetic Underwater Benchmark Dataset
Namritha Lasyapriya Maddali, Rajini Makam, Suresh Sundaram, Narasimhan Sundararajan
Submitted to Transactions on Image Processing, IEEE 

A physics-informed synthetic underwater benchmark of 50,000+ paired degraded–reference images, generated using a modified Jaffe–McGlamery model spanning all ten Jerlov water types with depth-dependent irradiance and biological optical effects. It achieves a 46% lower FID than the best competing synthetic dataset, with trained models improving UIQM by 4.2%, reducing NIQE by 23.9%, and nearly doubling keypoint matches over the strongest synthetic baselines.

MOVE-IT: Motion-Varying Input for Enhanced Temporal Integrity in Next-Frame Prediction
Yashas HG*, Namritha Maddali*, Chandana MS, Dhanush M, Aditya NG, Prasad Honnavalli, Gowri Srinivasa
11th ICICT'26, London, UK (scopus, el compendex indexed, Springer LNNS)

A motion-aware preprocessing method that filters low-motion sequences and adapts window size, achieving 2–3.5 dB PSNR and 7–15% SSIM gains while reducing training data by 50–65%.

Mind the Switch: Predicting Code-Switch Boundaries
Namritha Lasyapriya Maddali, Uma D, Skanda Sheersha Prasad
20th iSAI-NLP'25, Phuket, Thailand (scopus-indexed)

Created a QA Kanglish dataset, and a framework to annotate and predict the code-switching in these sentences, achieving an accuracy of about 90%