# Pseudo-Label Refinement for Robust Wheat Head Segmentation via Two-Stage Hybrid Training

Source: [Pseudo-Label Refinement for Robust Wheat Head Segmentation via Two-Stage Hybrid Training](<https://www.sota2.com/research/paper/6940e791456146936284f74d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-12-17T05:03:45.767Z

Authors: Jiahao Jiang, Zhangrui Yang, Xuanhan Wang, Jingkuan Song

Published: 2025-12-07T02:48:50.000Z

## Abstract

This extended abstract details our solution for the Global Wheat Full Semantic Segmentation Competition. We developed a systematic self-training framework. This framework combines a two-stage hybrid training strategy with extensive data augmentation. Our core model is SegFormer with a Mix Transformer (MiT-B4) backbone. We employ an iterative teacher-student loop. This loop progressively refines model accuracy. It also maximizes data utilization. Our method achieved competitive performance. This was evident on both the Development and Testing Phase datasets.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Semantic Segmentation on Global Wheat Full (dev)](<https://www.sota2.com/research/sota/semantic-segmentation-on-global-wheat-full-dev.md>): Semantic segmentation; Global Wheat Full (dev); mIoU: 0.748.

- [Semantic Segmentation on GWFSS 1.0 (test)](<https://www.sota2.com/research/sota/semantic-segmentation-on-gwfss-1-0-test.md>): Semantic segmentation; GWFSS 1.0 (test); mIoU: 70.99.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2512.11874>)

[PDF](<https://arxiv.org/pdf/2512.11874.pdf>)