Uncertain causes of guitar stand underperformance

Ascertain the specific factors responsible for the weak segmentation performance of the MVImgNet guitar stand class under viewpoint shifts in the Hummingbird-based multi-view evaluation.

Background

The guitar stand class performs well below baseline across models, with steep degradation under harder conditions. The thin structure is suspected to contribute to failure, but the authors explicitly state that the exact cause is unknown.

Identifying the cause would inform whether model design, training objectives, or dataset issues are the primary drivers of poor generalization for this class.

References

However, exact cause remains unclear and is a point for future investigation.

Evaluating Foundation Models' 3D Understanding Through Multi-View Correspondence Analysis (2512.11574 - Lilova et al., 12 Dec 2025) in Appendix A, Experiment A: guitar stand figure caption