ConvFaceNeXt: Lightweight Networks for Face Recognition

Hoo, Seng Chun; Ibrahim, Haidi; Suandi, Shahrel Azmin

doi:10.3390/math10193592

Cited by 6 publications

(2 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The Inverted Residual Block (IRB) [16] is a variation of the residual network commonly applied in image processing tasks. It usually converts low-resolution images into highresolution images, and has a good ability to restore lost detail information and improve clarity.…”

Section: B Double-layer Inverted Residual Blockmentioning

confidence: 99%

DIRBW-Net: An Improved Inverted Residual Network Model for Underwater Image Enhancement

An,

Feng,

Yuan

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Underwater photography is challenged by optical distortions caused by water's absorption and scattering phenomena. These distortions manifest as color aberrations, image blurring, and reduced contrast in underwater scenes. To address these issues, we propose an underwater image enhancement model leveraging an inverted residual network. Firstly, the traditional inverted residual network is improved. In order to minimize the interference of the Batch Normalization (BN) layer on color information, a novel Double-layer Inverted Residual Block (DIRB) is introduced, which omitted the BN layer and extracted deep feature information from the input image. Subsequently, the input image is fused with the intermediate feature map using a skip connection to ensure the consistency between local and global image information, thus effectively enhancing the image quality. In the concluding phase, we delved into the effects of diverse activation functions on the network model's performance, opting for the h-swish activation function to further boost the overall performance. The model underwent evaluation and testing on a public dataset, with comparisons drawn against representative enhancement models. The experiments showcase a notable success in enhancing the underwater image quality when using the proposed model.INDEX TERMS underwater image enhancement, convolutional neural network (CNN), residual network, deep learning.

show abstract

Section: B Double-layer Inverted Residual Blockmentioning

confidence: 99%

DIRBW-Net: An Improved Inverted Residual Network Model for Underwater Image Enhancement

An,

Feng,

Yuan

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…PocketNet, a student model, learns from the ResNet‐100 as its teacher model using the differential architecture search (DARTS) algorithm [29]. ConvFaceNet utilizes an enhanced ConvNeXt (ECN) block with depthwise convolution in the first layer [30] to achieve lower FLOPs and reduce the number of parameters. Martínez‐Díaz et al.…”

Section: Related Workmentioning

confidence: 99%

A lightweight deep learning model for real‐time face recognition

Deng

Chiang

Kang

et al. 2023

IET Image Processing

View full text Add to dashboard Cite

Lightweight deep learning models for face recognition are becoming increasingly crucial for deployment on resource‐constrained devices such as embedded systems or mobile devices. This paper presents a highly efficient and compact deep learning (DL) model that achieves state‐of‐the‐art performance on various face recognition benchmarks. The developed DL model employs one‐ or few‐shot learning to obtain effective feature embeddings and draws inspiration from FaceNet with significant refinements to achieve a memory size of only 3.5 MB—about 30 times smaller than FaceNet—while maintaining high accuracy and real‐time performance. The study demonstrates the model's effectiveness through extensive experiments, which include testing on public datasets and the model's ability to recognize occluded faces in uncontrolled environments using grayscale input images. Compared to the state‐of‐the‐art lightweight models, the proposed model requires fewer FLOPs (0.06G), has a smaller number of parameters (1.2 M), and occupies a smaller model size (3.5 MB) while achieving a competitive level of recognition accuracy and real‐time performance. The results show that the model is well‐suited for deployment in embedded domains, including live entrance security checks, driver authorization, and in‐class attendance systems. The entire code of FN8 is available on GitHub.

show abstract