MultiModal Sentence Summarization (MMSS) aims to generate a brief summary based on the given source sentence and its associated image. Previous studies on MMSS have achieved success by either selecting the task-relevant visual information or filtering out the task-irrelevant visual information to help the textual modality to generate the summary. However, enhancing from a single perspective usually introduces over-preservation or over-compression problems. To tackle these issues, we resort to Information Bottleneck (IB), which seeks to find a maximally compressed mapping of the input information that preserves as much information about the target as possible. Specifically, we propose a novel method, T(3), which adopts IB to balance the Trade-off between Task-relevant and Task-irrelevant visual information through the variational inference framework. In this way, the task-irrelevant visual information is compressed to the utmost while the task-relevant visual information is maximally retained. With the holistic perspective, the generated summary could maintain as many key elements as possible while discarding the unnecessary ones as far as possible. Extensive experiments on the representative MMSS dataset demonstrate the superiority of our proposed method. Our code is available at https://github.com/YuanMinghuan/T3.

Coastal wetland ecosystems are among the most productive yet highly threatened systems in the world, and population growth and increasing economic development have resulted in extremely rapid degradation and loss of coastal wetlands. Synthetic aperture radar (SAR) has proved to be of great potential in wetland applications, such as characterizing wetland types and mapping wetland inundation extent, based on the fact that the SAR backscatter signal from the wetland depends mainly on the wetland vegetation and hydrodynamic. However, the absence of ground observations has limited calibration and validation of water depth estimated from satellite SAR data. This letter aims to explore the potential for retrieving the water depth in freshwater marshes using L-band SAR backscattering coefficients. We present an innovative approach for combining backscattering coefficients and interferometric data to estimate water depth from L-band SAR data. The technique was applied to freshwater marshes in the Yellow River Delta using an Advanced Land Observing Satellite Phased Array L-band SAR data set acquired between 2007 and 2010. It is shown that the combination of L-band backscattering coefficients and interferometric data could provide information on the hydrodynamic of the Yellow River Delta with high spatial resolution.