Subspace methods help detect real defects from synthetic training in 3d inspection
Subspace Representations for Synthetic-to-Real Transfer in 3-D Surface Inspection
Computer Vision and Pattern RecognitionMachine Learning
Summary
Detecting surface defects on 3D scanned objects usually needs many real defect examples. This paper looks at how using simplified mathematical spaces called subspaces can help transfer learning from fake defect data to real defects. The authors find that the effectiveness depends on how well these subspaces cover the differences between the fake and real data. They also show that using a good reference scan and aligning the objects properly matters more than the type of subspace used.
What this means in practice
- •For manufacturing quality teams: Improve detection of real surface defects using models trained on synthetic defects with only good part scans available beforehand.
- •For robotics perception engineers: Use reference scans and registration in 3D inspection systems to enhance defect detection when real defect data is scarce.
Authors
Abdulla Nasir Chowdhury, Naish Ibrahim, Gulpi Qorik Oktagalu Pratamasunu, Kazuhiro Fukui
Abstract
This paper studies what subspace representations contribute to synthetic-to-real transfer in 3-D surface inspection, where models learn from synthetic defects and the only real data before deployment are scans of good parts. We use three classical constructions as instruments: the nearest-subspace classifier, one-class template subspaces and the generalized difference subspace (GDS). On 1,197 MVTec 3D-AD scans without target annotation we find that every decision rule ranks real defects at a similar window AUC, that image-level detection stays at chance, and that no rule keeps its in-domain threshold. We show that the margin of a nearest-subspace rule moves by at most the part of the feature shift outside its class subspaces, and that this leakage, measured from unlabeled good scans, bounds the displacement observed on unseen scans. On 597 Real3D-AD objects, a one-class subspace of registered templates evaluated at every location on frozen features reaches point AUROC 0.957 on pre-registered scans and a category-mean object AUROC of 78 to 81 with no training on the benchmark. Our results suggest that a subspace helps under shift in proportion to the shift it spans, and that in reference-based inspection the reference and its registration matter more than the subspace.