Embedding-Only Uplink for Onboard Retrieval Under Shift in Remote Sensing
For remote sensing systems with downlink bottlenecks, this work demonstrates that a single set of uplinked embeddings can support multiple tasks with task-dependent optimal heads, enabling efficient onboard triage without additional uplink cost.
The paper investigates an embedding-only uplink pipeline for onboard remote sensing retrieval under distribution shift, finding that kNN retrieval outperforms centroid methods for cloud classification (0.92 vs. 0.91) while centroids dominate for temporal change detection (0.85 vs. 0.48), with all telemetry under 1 KB per query.
Downlink bottlenecks motivate onboard systems that prioritize hazards without transmitting raw pixels. We study a strict setting where a ground station uplinks only compact embeddings plus metadata, and an onboard system performs vector search to triage new captures. We ask whether this embedding-only pipeline remains useful under explicit remote-sensing shift: cross-time (pre/post-event), cross-event/location (different disasters), cross-site cloud (15 geographic sites), and cross-city AOI holdout (buildings). Using OlmoEarth embeddings on a scaled public multi-task benchmark (27 Sentinel-2 L2A scenes, 15 cloud sites, 5 SpaceNet-2 AOIs; 10 seeds), we find that all effective methods rely on the same uplinked embeddings, but the optimal decision head is task-dependent: kNN retrieval is significantly superior for cloud classification (0.92 vs. centroid 0.91; p<0.01, Wilcoxon), while class centroids dominate temporal change detection (0.85 vs. retrieval 0.48; p<0.01). These results show that embedding-only uplink is the key enabler--once embeddings are onboard, the system can select the best head per task at no additional uplink cost, with all telemetry under 1 KB per query.