ProGuT: Label-Efficient Panoptic Segmentation for Forest Scenes

2026-09-29

AI Summary

This study addresses the challenges of low-quality instance segmentation and reliance on expensive sensors in forest scenes by proposing an unsupervised panoptic segmentation framework. The method generates pseudo-labels through CLIP feature clustering combined with multi-scale geometric priors, thereby eliminating the need for per-image annotation. Furthermore, a prototype-guided training strategy is designed, incorporating structure tensor analysis to suppress background interference and accurately recover tree trunk instances. Experimental results demonstrate that the proposed model achieves a Panoptic Quality (PQ) of 65.2 and a mean Intersection over Union (mIoU) of 65.9, significantly outperforming baseline methods such as PiCIE. This work thus enables cost-effective, high-quality understanding of complex forest environments.

Abstract

Panoptic segmentation in forest environments is bottlenecked not by semantic quality but by instance separation; existing unsupervised panoptic approaches produce usable stuff maps but near-zero thing quality. Depth or flow-based instance discovery methods needs sensors that are not always available. We present ProGuT (Prototype Guided Training), which produces panoptic pseudo-labels without per-image training masks, needing only unlabeled images and one-time cluster-to-class mapping. ProGuT clusters CLIP patch features, then recovers trunk instances through multiscale geometric prior that falsifies non-trunk structures via structure-tensor. This is cheap compared to depth, flow or class-supervision methods to create pseudo labels. These are then used for downstream tasks which we evaluate against other unsupervised baselines. ProGuT achieves a Panoptic Quality (PQ) of 65.2 on Our-forest dataset (2.6x improvement over the initial pseudo-label quality) and reaches 65.9 mIoU on Freiburg Forest, outperforming unsupervised baselines like PiCIE (45.3 IoU) and STEGO(57.6IoU). Additionally, ProGuT outperforms existing unsupervised methods for class-agnostic trunk instance benchmark.

Problem

• Panoptic Segmentation
• Forest Scenes
• Instance Separation
• Unsupervised Learning
• Label Efficiency

Innovation

• Panoptic Segmentation
• Prototype Guided Training
• CLIP Patch Features
• Structure-Tensor
• Label-Efficient