# QFacePlacement: From Arbitrary Tensorization to Semantic Spatial Correspondence in Modern Face Verification

DevFeed: [QFacePlacement: From Arbitrary Tensorization to Semantic Spatial Correspondence in Modern Face Verification](<https://devfeed.tech/articles/qfaceplacement-from-arbitrary-tensorization-to-semantic-spatial-correspondence-in-modern-face-verification-114714.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.02505>)

Author: Krishna Bhatia, Kavita Choudhary, Pruthwik Mishra, Praveen Kumar Chandaliya

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Facial recognition](<https://devfeed.tech/topics/facial-recognition.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>)

Tags: [alignment](<https://devfeed.tech/tags/alignment.md>), [compression](<https://devfeed.tech/tags/compression.md>), [computer-vision](<https://devfeed.tech/tags/computer-vision.md>), [cosine](<https://devfeed.tech/tags/cosine.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [features](<https://devfeed.tech/tags/features.md>), [gallery](<https://devfeed.tech/tags/gallery.md>), [global](<https://devfeed.tech/tags/global.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [linear](<https://devfeed.tech/tags/linear.md>), [local](<https://devfeed.tech/tags/local.md>), [matching](<https://devfeed.tech/tags/matching.md>), [matrix](<https://devfeed.tech/tags/matrix.md>), [metrics](<https://devfeed.tech/tags/metrics.md>), [model](<https://devfeed.tech/tags/model.md>), [mpo](<https://devfeed.tech/tags/mpo.md>), [quantum](<https://devfeed.tech/tags/quantum.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [verification](<https://devfeed.tech/tags/verification.md>)

## AI overview

The study finds that tensor compression works better when applied to semantically aligned spatial correspondences than to arbitrary face-embedding reshapes, but compactness does not ensure strong verification at low false-accept rates. For mapping 512-dimensional AdaFace embeddings into ArcFace space, dense Ridge performs better than compressed alternatives at a 64K-parameter budget: mean directional AUC is 0.9297 for Ridge, 0.8952 for rank-64 SVD, 0.8685 for the best low-rank MPO hybrid, and about 0.649 for raw MPO. A geometry-aware symmetric tensor train applied to local descriptor correspondences uses 122 parameters versus 2,403 for a dense linear matcher and reaches 0.9802 LFW AUC. However, its TAR at FAR 0.001 is lower than global cosine matching on CFP-FP (0.0102 vs. 0.1734) and CPLFW (0.7805 vs. 0.8737), underscoring the need to assess operational low-FAR metrics alongside aggregate AUC.

## Source excerpt

arXiv:2611.02505v1 Announce Type: cross Abstract: Quantum-inspired computer vision often reshapes high-dimensional features into tensors and applies tensor-network models. In face verification, this can be misleading: modern embeddings are optimized for cosine similarity, and tensorization alone does not establish the low-rank structure needed for tensor trains or matrix-product operators (MPOs). We study model placement through two controlled interfaces. First, cross-model embedding alignment requires a 512 x 512 operator to map AdaFace embeddings into ArcFace space. Dense Ridge and Procrustes mappings remain strong, while deterministic TT-SVD and MPO compression underperform even with basis changes and low-rank residuals. At a 64K-parameter storage budget, mean directional AUC is 0.9297 for dense Ridge, 0.8952 for rank-64 SVD, 0.8685 for the best low-rank-MPO hybrid, and approximately 0.649 for raw MPO. Second, before global pooling, we construct a 7 x 7 x 7 x 7 probe-gallery correspondence tensor from local AdaFace descriptors. A geometry-aware symmetric tensor train with bond dimension 2 uses 122 trainable parameters, compared with 2,403 for a dense linear matcher, while attaining an LFW AUC of 0.9802. Yet compression does not translate into better low-false-acceptance performance: on CFP-FP, TAR at FAR = 0.001 is 0.0102 versus 0.1734 for global cosine matching, and on CPLFW it is 0.7805 versus 0.8737. These results suggest that quantum-inspired factorization is most credible when the chosen interface provides task-aligned axes and compressible structure. They also show why aggregate AUC must be assessed alongside operational low-FAR metrics.