# On the Capacity of Face Representation

arXiv: https://arxiv.org/abs/1709.10433 · HTML: https://mlsanity.com/paper/1709.10433/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 29 Sep 2017.

## Built on by (most cited first, 5 of 5)

- [ReliOcc: Towards Reliable Semantic Occupancy Prediction via Uncertainty Learning](https://mlsanity.com/paper/2409.18026/) (2024)
- [On the Biometric Capacity of Generative Face Models](https://mlsanity.com/paper/2308.02065/) (2023)
- [Multi-Biometric Fuzzy Vault based on Face and Fingerprints](https://mlsanity.com/paper/2301.06882/) (2023)
- [Understanding Biometric Entropy and Iris Capacity: Avoiding Identity Collisions on National Scales](https://mlsanity.com/paper/2308.03189/) (2023)
- [Empirical Assessment of End-to-End Iris Recognition System Capacity](https://mlsanity.com/paper/2303.12742/) (2023)

## Abstract (arXiv)

In this paper we address the following question, given a face representation, how many identities can it resolve? In other words, what is the capacity of the face representation? A scientific basis for estimating the capacity of a given face representation will not only benefit the evaluation and comparison of different representation methods, but will also establish an upper bound on the scalability of an automatic face recognition system. We cast the face capacity problem in terms of packing bounds on a low-dimensional manifold embedded within a deep representation space. By explicitly accounting for the manifold structure of the representation as well two different sources of representational noise: epistemic (model) uncertainty and aleatoric (data) variability, our approach is able to estimate the capacity of a given face representation. To demonstrate the efficacy of our approach, we estimate the capacity of two deep neural network based face representations, namely 128-dimensional FaceNet and 512-dimensional SphereFace. Numerical experiments on unconstrained faces (IJB-C) provides a capacity upper bound of 2.7×104 for FaceNet and 8.4×104 for SphereFace representation at a false acceptance rate (FAR) of 1%. As expected, capacity reduces drastically at lower FARs. The capacity at FAR of 0.1% and 0.001% is 2.2×103 and 1.6×101, respectively for FaceNet and 3.6×103 and 6.0×100, respectively for SphereFace.