海角社区

Changes in technology are inevitable. However, while some advances are transitory, others make a lasting impact, standing the test of time.

In 2015, DES Professor Supratik Mukhopadhyay and a team of researchers published a paper on , a novel, specialized toolbox designed to improve understanding of one of the most common types of remote sensing data 鈥� satellite imagery.

Now, a decade later, the impact of that research has been recognized as Mukhopadhyay and his team were runners up in the Association for Computing Machinery, or ACM鈥檚, SIGSPATIAL 10-Year Impact Award, an award given out to recognize research that has continued to contribute over time.  

The DeepSat framework improved systems鈥� abilities to analyze satellite images, by utilizing deep learning combined with high resolution data to improve classification systems.  

Over the next decade, the technique drove an 11-15 percent leap in accuracy in high resolution satellite imagery classification.

It has been cited approximately 489 times in Google Scholar and, according to SIGSPATIAL, has about 20 active downloads a month.

Not only that, it has been actively used in research applications, including by NASA鈥檚 Earth science programs, where it has been used in landcover mapping and carbon monitoring. It was also used to develop a landcover map of California at a one-meter resolution.

鈥淭his recognition means a great deal because it reflects sustained influence, not just a moment of novelty, but a contribution that continued to shape research and real-world applications over a decade. DeepSat was one of the early works to demonstrate how deep learning could be effectively adapted for understanding high resolution satellite imagery, before high res. satellite-derived products and geospatial AI became mainstream in climate, disaster resilience, agriculture, wildfire monitoring, Carbon capture, and planetary-scale analysis,鈥� Mukhopadhyay said.

 

---