Granularity-Aware Transfer for Tree Instance Segmentation in Synthetic and Real Forests

TL;DR

This paper proposes a granularity-aware transfer method for tree instance segmentation, leveraging synthetic fine-grained data to improve coarse real-world labels, with a new dataset and distillation protocol.

Contribution

Introduction of MGTD, a mixed-granularity dataset, and a novel distillation method that transfers structural priors from synthetic to real data.

Findings

Consistent mask AP improvements, especially for small/distant trees.

Establishment of a new testbed for Sim-Real transfer under label granularity constraints.

Abstract

We address the challenge of synthetic-to-real transfer in forestry perception where real data have only coarse Tree labels while synthetic data provide fine-grained trunk/crown annotations. We introduce MGTD, a mixed-granularity dataset with 53k synthetic and 3.6k real images, and a four-stage protocol isolating domain shift and granularity mismatch. Our core contribution is granularity-aware distillation, which transfers structural priors from fine-grained synthetic teachers to a coarse-label student via logit-space merging and mask unification. Experiments show consistent mask AP gains, especially for small/distant trees, establishing a testbed for Sim-Real transfer under label granularity constraints.