Efficient crop mapping from satellite images using smaller smart models
JEDI: JEPA-to-Edge Distillation for Efficient Cropland Segmentation from Satellite Imagery
Computer Vision and Pattern RecognitionMachine Learning
Summary
Large computer vision models can help identify different types of farmland from satellite images, but they are often too big and slow to use directly on satellites or small devices. The authors present a method called JEDI that teaches a smaller, faster model to mimic a very large model’s understanding without losing much accuracy. They do this by carefully matching the small model’s internal data to the big model’s, even while the small model learns its actual task. Their approach works well on multiple sizes of small models, significantly improving performance while keeping the model size and computing needs much lower.
vision transformerfeature distillationsegmentationremote sensingmodel compressionrepresentation learningstudent-teacher modelcross-architecture alignmentsatellite imagerymachine learning
Authors
Kishor Kumar Bhaumik, Nicolas Roque dos Santos, Jia Chen, Evangelos E. Papalexakis
Abstract
Large vision models provide useful representations for remote-sensing segmentation but are often too expensive for deployment at the satellite or field edge. Existing feature-level distillation methods also tend to assume similar teacher and student architectures and often stop feature alignment when task training begins. We introduce JEDI (JEPA-to-Edge Distillation), a two-stage framework that transfers representations from a large I-JEPA Vision Transformer teacher to a compact SegFormer student. First, JEDI aligns the student's terminal representation with the teacher's token space using cross-architecture projection and spatial alignment. It then jointly optimizes supervised segmentation, temperature-scaled response distillation, and persistent feature alignment throughout task adaptation. On CalCROP21, JEDI-B0 achieves 68.0 mean Intersection-over-Union (mIoU) with 4.04M parameters, improving over the standalone student by 16.0 points and coming within 2.0 points of the 70.0 mIoU achieved by the 639M-parameter teacher. We evaluate SegFormer B0, B1, and B2 students with 4.04M, 14.33M, and 28M parameters, respectively. Across all three variants, JEDI consistently outperforms response-, structure-, channel-, and relational-distillation baselines under the same teacher-student setting. These results show that persistent representation alignment is especially valuable under aggressive compression, substantially reducing model size and computation while preserving segmentation performance.