Compact encoder restores satellite images efficiently without attention
LIMODENet: Attention-Free Compact Encoders for Information-Preserving Onboard Satellite Image Restoration
Computer Vision and Pattern Recognition
Summary
Satellites take images that can get degraded when sent back to Earth. Fixing these images onboard is hard because satellites have limited power and special processors that don’t support common AI tricks like attention. The authors created LIMODENet, a small neural network that restores images well without using attention. It keeps more information through its layers and works better than some existing smaller models under these hardware limits. While bigger models can restore images even better, they can’t run on these low-power satellite systems.
What this means in practice
- •For satellite system engineers: Design onboard image restoration software for satellites using neuromorphic chips without attention support.
- •For embedded ai developers: Build efficient neural networks that restore degraded images on power-constrained devices lacking attention mechanisms.
Authors
Thanh-Dung Le, Vu Nguyen Ha, Ti Ti Nguyen, Symeon Chatzinotas
Abstract
Onboard satellites must restore a channel-degraded image on a few watts, using neuromorphic accelerators (e.g., BrainChip Akida, Intel Loihi-2) that support no softmax or attention. We ask which encoder restores best under that constraint and introduce LIMODENet (LinearMix-ODENet), a 0.69M softmax-/QKV-free backbone whose residual stages read as ODE discretizations and which is empirically information-preserving (probe accuracy rises 79.9% -> 98.4% from stem to head). At iso-parameters it restores 1 dB DVB-S2X-degraded EuroSAT better than a CNN autoencoder (+1.75 dB PSNR) and a skip-connection U-Net (+1.07 dB), three seeds, non-overlapping. Unconstrained modern restorers (NAFNet, Restormer) win on fidelity; we decompose that gap: spiking-legal additive skips recover about half, and the rest traces to attention and channel gating. LIMODENet then converts end-to-end to a spiking network with zero blocked operations, versus 22-24 for the competitors: not the best restorer available, but the best verified deployable within a real power budget.