Weakly Supervised Semantic Segmentation For Remote Sensing Hyperspectral Imaging

A novel approach to semantic segmentation using weakly supervised learning significantly outperforms traditional methods, such as Support Vector Machines (SVM), when applied to hyperspectral imaging data from Teide National Park. The proposed Deep Seeded Region Growing system leverages a neural network trained on seeds generated by an SVM, progressively increasing pixel-level supervision through a region-growing algorithm. This method enhances the accuracy and efficiency of vegetation mapping in remote sensing applications, offering a promising alternative for large-scale environmental monitoring.

Moliner, E., Romero, L. S., & Vilaplana, V. (2020). Weakly Supervised Semantic Segmentation For Remote Sensing Hyperspectral Imaging. ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2273–2277. https://doi.org/10.1109/icassp40776.2020.9053384