Quantum-inspired methods struggle to improve face verification matching
QFacePlacement: From Arbitrary Tensorization to Semantic Spatial Correspondence in Modern Face Verification
Computer Vision and Pattern RecognitionMachine Learning
Summary
Face verification systems often compare face images by turning features into multi-dimensional structures called tensors. This paper finds that simply reshaping features into tensors without careful alignment doesn’t improve matching accuracy and can even worsen results. The authors show that traditional dense methods perform better than quantum-inspired low-rank tensor methods, especially in tasks requiring very low false acceptance rates. Their work highlights that using quantum-inspired models is only beneficial when the data is well aligned to the model's structure.
What this means in practice
- •For facial recognition engineers: Design better embedding alignment systems by preferring dense linear mappings over quantum-inspired low-rank tensor factorizations for improved verification accuracy.
- •For security teams using biometric controls: Understand the limits of tensor-based compression methods to avoid degraded low-false-acceptance rate performance in biometric access systems.
Authors
Krishna Bhatia, Kavita Choudhary, Pruthwik Mishra, Praveen Kumar Chandaliya
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.