PARCS: A Deployment-Oriented AI System for Robust Parcel-Level Cropland Segmentation of Satellite Images

Cropland segmentation of satellite images is an essential basis for crop area and yield estimation tasks in the remote sensing computer vision interdisciplinary community. Instead common pixel-level results with salt-and-pepper effects, a parcel-level output conforming to human recognition required according clients' needs during model deployment. However, leveraging CNN-based models requires fine-grained labels, which unacceptable annotation burden. To cure these practical pain points, this paper, we present PARCS, holistic deployment-oriented AI system PARcel-level Segmentation. By consolidating multi-disciplinary knowledge, PARCS has two algorithm branches. The first branch performs by learning from limited labeled pixel samples active strategy avoid costs. second aims at generating parcel regions without procedure. final result achieved integrating outputs branches tandem. robust effectiveness demonstrated its outstanding performance on public in-house datasets (an overall accuracy 85.3% mIoU 61.7% PASTIS dataset, 65.16% dataset). We also include subjective feedback clients discuss lessons learned