This study introduces a approach to cross-age facial recognition, highlighting the crucial role of extracting rich hybrid features and isolating identity characteristics within them. Our proposed method integrates a Feature Aggregation and Selection Module with a Non-linear Identity Feature Separation module. The process begins with the generation of hybrid features through an attention-driven fusion of basic and advanced semantic features. This is followed by the extraction of identity features via Non-linear decoupling, guided by multi-task training. These identity features are then applied to cross-age facial recognition tasks. The effectiveness and adaptability of our approach are demonstrated by its impressive performance on various datasets, including AgeDB-30, CALFW, CACD-VS, and LFW, achieving accuracy rates of 97.10%, 96.22%, 99.61%, and 99.71% respectively. This method represents a significant advancement in the field of facial recognition research, particularly in addressing the challenges of age variation.

This paper introduces a lightweight spectral channel feature transformation network, LDCFormer, designed to address the high computational complexity and excessive parameter count resulting from self-attention and spatial MLP (Multilayer Perceptron) in vision transformers when dealing with long sequences. Initially, the image information is transformed into the frequency domain using a two-dimensional discrete cosine transform (DCT), effectively capturing the image’s frequency domain features. Secondly, considering that different frequency areas represent various types of features, local feature information such as edges and textures are extracted in the high-frequency area, while the image’s global feature information is extracted in the low-frequency area, complemented by channel attention for feature cleansing. Finally, the integration and interaction of global feature information in the image are achieved by introducing the Transformer architecture. LDCFormer employs a zero-learning-parameter 2D LDCFormer operation to extract features directly from the frequency domain, significantly reducing the number of trainable parameters, and utilizes depth-separable LDConv MLP to further accelerate computational speed, achieving the lightweight and efficient characteristics of LDCFormer. Accuracies of 78.8%, 88.9% and 88.6% were attained on three typical datasets. The experimental results demonstrate that LDCFormer maintains high classification performance while reducing the parameter count, achieving a good balance between speed and accuracy.

The construction of deep neural networks depends on a significant number of parameters and computational complexity, which poses a challenge in the field of image processing. To address the issue of the Transformer network model's large size and inability to effectively capture local features of the image, this paper proposes a lightweight composite Transformer structure that combines a spectral feature refinement module (SFRM) and a parameterless attention augmentation module (PAAM). The SFRM and PAAM work together to improve the quality of the spectral features used in the transformer. The proposed structure aims to enhance the performance of the transformer without adding unnecessary complexity. The SFRM utilises the two-dimensional discrete cosine transform to convert the image from the spatial domain to the frequency domain. This process extracts both the overall image structure and detailed feature information from the high-frequency and low-frequency regions, respectively. The aim is to purify the spatially-insignificant features in the original image. The PAAM introduces a parameter-free channel, spatial, and 3D attention enhancement mechanism to extract correlation features of local information in the spatial domain without increasing the number of parameters. This improves the expression of local features in the image. Additionally, Depth Separable (DConv MLP) is introduced to further reduce the network model's weight. The experimental results show that the proposed algorithm achieves an accuracy of 79.6% on the ImageNet-1K dataset, 91.6% on the Oxford 102 Flower Dataset, and 94.1% on the CIFAR-10 dataset. Compared to ViT-B, Swin-T, and CSwin-T, respectively, the number of covariates decreases by 86.11%, 58.62%, and 47.83%. The number of parameters is also lower than VGG-16 and ResNet-110 by 91.07% and 77.70%, respectively.