Yadav Raj Ghimire

Hybrid FLAVA achieves superior multimodal reasoning in Remote Sensing VQA with less than 5% of trainable parameters, redefining efficiency in the field.

Dual-domain deep learning models not only outperform single-domain counterparts in disaster damage assessment but also reveal critical limitations in detecting subtle damage levels.

U-Net++ with EfficientNet-B3 outperformed other architectures, achieving an impressive 97.62% mIoU for rice disease mapping, underscoring the effectiveness of lightweight models in agricultural monitoring.