Very excited to be flying to Denver next week for the upcoming CVPR conference!
I will be presenting our latest paper on embedded image restoration for satellite imagery! In this work, with my co-workers Marjorie Bellizzi and Omar HLIMI, we show that there are efficient ways to improve the information contained in a noisy image that can be used for real-time decision-making directly in space. We benchmarked our method via an object detection use case, and the results are really promising!
As satellites become smarter and more autonomous, we envision that making better use of imperfect, noisy information will be key to enabling faster and more reliable decisions onboard.
Looking forward to sharing our work and connecting with the community!
For the curious, the paper is available here:
[2604.12807] Rethinking Satellite Image Restoration for Onboard AI: A Lightweight Learning-Based Approach