SG-DOR: Learning Scene Graphs with Direction-Conditioned Occlusion Reasoning for Pepper Plants
This work addresses the challenge of effective robotic harvesting in agriculture by providing a structured relational signal for intervention planning, though it is incremental as it builds on existing scene graph and occlusion reasoning methods.
The paper tackles the problem of robotic harvesting in dense crop canopies by developing a relational framework, SG-DOR, that infers scene graphs to encode physical attachments and direction-conditioned occlusion for pepper plants, resulting in improved occlusion prediction (F1=0.73, NDCG@3=0.85) and attachment inference (edge F1=0.83) over ablations.
Robotic harvesting in dense crop canopies requires effective interventions that depend not only on geometry, but also on explicit, direction-conditioned relations identifying which organs obstruct a target fruit. We present SG-DOR (Scene Graphs with Direction-Conditioned Occlusion Reasoning), a relational framework that, given instance-segmented organ point clouds, infers a scene graph encoding physical attachments and direction-conditioned occlusion. We introduce an occlusion ranking task for retrieving and ranking candidate leaves for a target fruit and approach direction, and propose a direction-aware graph neural architecture with per-fruit leaf-set attention and union-level aggregation. Experiments on a multi-plant synthetic pepper dataset show improved occlusion prediction (F1=0.73, NDCG@3=0.85) and attachment inference (edge F1=0.83) over strong ablations, yielding a structured relational signal for downstream intervention planning.