CoChat

Face Recognition Based on Deep Learning and FPGA for Ethnicity Identification

Open access

The DOI matches its Crossref record; 52 of the 52 references it cites were found in a registry or an index; 8 of the 16 sources its claims cite were checked for support.

Abstract

In the last decade, there has been a surge of interest in addressing complex Computer Vision (CV) problems in the field of face recognition (FR). In particular, one of the most difficult ones is based on the accurate determination of the ethnicity of mankind. In this regard, a new classification method using Machine Learning (ML) tools is proposed in this paper. Specifically, a new Deep Learning (DL) approach based on a Deep Convolutional Neural Network (DCNN) model is developed, which outperforms a reliable determination of the ethnicity of people based on their facial features. However, it is necessary to make use of specialized high-performance computing (HPC) hardware to build a workable DCNN-based FR system due to the low computation power given by the current central processing units (CPUs). Recently, the latter approach has increased the efficiency of the network in terms of power usage and execution time. Then, the usage of field-programmable gate arrays (FPGAs) was considered in this work. The performance of the new DCNN-based FR method using FPGA was compared against that using graphics processing units (GPUs). The experimental results considered an image dataset composed of 3141 photographs of citizens from three distinct countries. To our knowledge, this is the first image collection gathered specifically to address the ethnicity identification problem. Additionally, the ethnicity dataset was made publicly available as a novel contribution to this work. Finally, the experimental results proved the high performance provided by the proposed DCNN model using FPGAs, achieving an accuracy level of 96.9 percent and an F1 score of 94.6 percent while using a reasonable amount of energy and hardware resources.

Overview

CoChat Summary

TL;DR

The paper proposes a deep convolutional neural network for ethnicity identification from facial images and implements it on an FPGA DE1-SoC, comparing it with GPU execution and four pre-trained CNNs. It reports 96.9% accuracy and 94.6% F1 on a new 3,141-image dataset of Chinese, Pakistani, and Russian faces.

Key findings

  • The proposed network achieved accuracy 0.969 and F1 score 0.946, outperforming ResNet50 (0.949 accuracy, 0.924 F1), DenseNet (0.957, 0.934), AlexNet (0.879, 0.799), and GoogleNet (0.940, 0.904) in Table 2.
  • On the FPGA DE1-SoC, the proposed network’s execution time and power consumption were reported as 2.76 s and 10.7 W, compared with 5.21 s and 650 W on GPU (Table 3).
  • The new dataset contained 3,141 face photographs: 1,081 Chinese, 1,021 Pakistani, and 1,039 Russian images, split into 60% training, 10% validation, and 30% test.
  • The FPGA implementation used 25,671 total logic elements, 96,734 total block memory bits, and 64,523 total logic registers (Table 4).

Methods

  • A new DCNN proposal with twelve deep learning layers: four convolutional layers each followed by max-pooling, dropout layers, two fully connected layers, a drop-connect layer, a flattening layer, and a SoftMax output layer; input images are 128 × 128 × 3 with 3 × 3 patches and stride 1.
  • The dataset was collected from three countries and compared against four pre-trained CNN models—ResNet50, DenseNet, AlexNet, and GoogleNet—with the last four layers of each frozen; training used an Intel Core i7-10750H CPU, NVIDIA GeForce RTX 2070 GPU, and 16 GB RAM.
  • FPGA acceleration used an FPGA DE1-SoC with a 64-computation array of 16-bit DSP; due to FPGA fabric limits, only 13 convolutional layers were employed. Evaluation included accuracy, F1, sensitivity, specificity, precision, false positive rate, execution time, power consumption, and DE1-SoC resource usage.

Limitations

  • The paper states that a direct numerical comparison against all state-of-the-art methods was not feasible because several were trained and assessed on inaccessible private datasets.
  • The paper states that only 13 convolutional layers were used because of the limits of FPGA fabrics on DE1-SoC.
  • Evident from the text: the dataset is limited to 3,141 images from three national/ethnic groups (Chinese, Pakistani, Russian), and no cross-dataset or broader demographic validation is reported, so generalization to other ethnicities or conditions is untested.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/app12052605

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Interconnection between PE (processing element) and S (switch). Figure 1. Interconnection between PE (processing element) and S (switch).
Figure 1 Interconnection between PE (processing element) and S (switch). Figure 1. Interconnection between PE (processing element) and S (switch).CC-BY · From the PDF
Figure 2 The block diagram of depth-wise convolution unit (DCU). Figure 2. The block diagram of depth-wise convolution unit (DCU).
Figure 2 The block diagram of depth-wise convolution unit (DCU). Figure 2. The block diagram of depth-wise convolution unit (DCU).CC-BY · From the PDF
Figure 3 Our proposal system for ethnicity identification based on deep learning. Figure 3. Our proposal system for ethnicity identification based on deep learning.
Figure 3 Our proposal system for ethnicity identification based on deep learning. Figure 3. Our proposal system for ethnicity identification based on deep learning.CC-BY · From the PDF
Image 4
CC-BY · From the PDF
Image 5
CC-BY · From the PDF
Image 6
CC-BY · From the PDF
Figure 4 A portion of the new dataset gathered from three separate countries. Figure 4. Aportion of the new dataset gathered from three separate countries.
Figure 4 A portion of the new dataset gathered from three separate countries. Figure 4. Aportion of the new dataset gathered from three separate countries.CC-BY · From the PDF
Image 8
CC-BY · From the PDF
Image 9
CC-BY · From the PDF
Image 10
CC-BY · From the PDF
Image 11
CC-BY · From the PDF
Image 12
CC-BY · From the PDF
Image 13
CC-BY · From the PDF
Image 14
CC-BY · From the PDF
Image 15
CC-BY · From the PDF
Figure 5 The schematic of a convolutional block within an FPGA. Figure 5. The schematic of a convolutional block within an FPGA.
Figure 5 The schematic of a convolutional block within an FPGA. Figure 5. The schematic of a convolutional block within an FPGA.CC-BY · From the PDF
Where: TP: True Positive, TN: True Negative, FP: False Positive and FN: False Negative. Where: TP: True Positive, TN: True Negative, FP: False Positive and FN: False Negative.
Where: TP: True Positive, TN: True Negative, FP: False Positive and FN: False Negative. Where: TP: True Positive, TN: True Negative, FP: False Positive and FN: False Negative.CC-BY · From the PDF
Image 18
CC-BY · From the PDF
Figure 7 ( A ) Time of execution ( B ) Power consumption. Figure 7. ( A ) Time of execution ( B ) Power consumption.
Figure 7 ( A ) Time of execution ( B ) Power consumption. Figure 7. ( A ) Time of execution ( B ) Power consumption.CC-BY · From the PDF
Figure 8 Numbers of resources for DE1-Soc. Figure 8. Numbers of resources for DE1-Soc.
Figure 8 Numbers of resources for DE1-Soc. Figure 8. Numbers of resources for DE1-Soc.CC-BY · From the PDF

Verification

Claims the paper makes

7 Valid · 4 Not valid · 7 Not checked · 34 Own results and interpretations

11 more claims

Create a free account to see all 52 claims and to verify the 3 references still open.

See all claims — free account
  • It may state two separate claims as one. 19
  • Probably not a substantial claim. 21
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Citations

Cites

Cited by

See all 54

Metadata

Authors
AlBdairi, A. J. A., Xiao, Z., Alkhayyat, A., Humaidi, A. J., Fadhel, M. A., Taher, B. H., Alzubaidi, L., Santamaría, J., & Al-Shamma, O.
Affiliations
  • Hunan University (CN)
  • Islamic University of Najaf (IQ)
  • University of Technology - Iraq (IQ)
  • University of Sumer (IQ)
  • Queensland University of Technology (AU)
  • Universidad de Jaén (ES)
  • University of Information Technology and Communications (IQ)
Corresponding author
  • Alzubaidi, Laith
  • Santamaría, José
Venue
Applied Sciences
Publisher
MDPI AG
Published
2022-03-02
Volume
12
Issue
5
Pages
2605
ISSN
2076-3417 (ISSN-L)
OpenAlex
W4214813523
Cited by
54
References
45
Citations by year
2026: 10 · 2025: 8 · 2024: 15 · 2023: 14 · 2022: 7
FWCI
3.60
Type
journal article
Language
en
Open access
Locations
  • Applied Sciences (publishedVersion, cc-by)
  • QUT ePrints (Queensland University of Technology) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face and Expression Recognition
Keywords
face recognition, deep convolutional neural networks, FPGA acceleration, GPU comparison, facial features, energy efficiency, hardware resource utilization

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.