Heterogeneous Face Recognition Research Articles

S2HConv-Caps-BiGRU: Deep Learning-Based Heterogeneous Face Recognition Model with Divergent Stages

Real-time images of faces captured in different spectrum bands are considered heterogeneous images. Heterogeneous Face Recognition (HFR) matches faces across domains and is crucial to public safety. This paper proposes an HFR approach based on Deep Neural Networks (DNN). Feature maps are extracted from two images, such as gallery and sketch images, using Squirrel Search Heterogeneous Convolutional-Capsule- Bidirectional Gated Recurrent Unit (S2HConv-Caps-BiGRU). As a method of efficiently recognizing faces, coupled representation similarity metric (CRSM) will use the measure for the similarity of two feature maps. The experimental results will be evaluated to state-of-the-art (SOTA) statistical measures in terms of accuracy, recall, Jaccard score, dice score, mean square error (MSE), image similarity, performance and root mean square error (RMSE). Compared to other SOTA, the model produces the best results. The accuracy value of a CUFS dataset is 98.7%.

A two-stage algorithm for heterogeneous face recognition using Deep Stacked PCA Descriptor (DSPD) and Coupled Discriminant Neighbourhood Embedding (CDNE)

A two-stage algorithm for heterogeneous face recognition using Deep Stacked PCA Descriptor (DSPD) and Coupled Discriminant Neighbourhood Embedding (CDNE)

On-the-move heterogeneous face recognition in frequency and spatial domain using sparse representation.

Heterogeneity of a probe image is one of the most complex challenges faced by researchers and implementers of current surveillance systems. This is due to existence of multiple cameras working in different spectral ranges in a single surveillance setup. This paper proposes two different approaches including spatial sparse representations (SSR) and frequency sparse representation (FSR) to recognize on-the-move heterogeneous face images with database of single sample per person (SSPP). SCface database, with five visual and two Infrared (IR) cameras, is taken as a benchmark for experiments, which is further confirmed using CASIA NIR-VIS 2.0 face database with 17580 visual and IR images. Similarity, comparison is performed for different scenarios such as, variation of distances from a camera and variation in sizes of face images and various visual and infrared (IR) modalities. Least square minimization based approach for finding the solution is used to match face images as it makes the recognition process simpler. A side by side comparison of both the proposed approaches with the state-of-the-art, classical, principal component analysis (PCA), kernel fisher analysis (KFA) and coupled kernel embedding (CKE) methods, along with modern low-rank preserving projection via graph regularized reconstruction (LRPP-GRR) method, is also presented. Experimental results suggest that the proposed approaches achieve superior performance.

From Modalities to Styles: Rethinking the Domain Gap in Heterogeneous Face Recognition

From Modalities to Styles: Rethinking the Domain Gap in Heterogeneous Face Recognition

Mind the Gap: Learning Modality-Agnostic Representations With a Cross-Modality UNet.

Cross-modality recognition has many important applications in science, law enforcement and entertainment. Popular methods to bridge the modality gap include reducing the distributional differences of representations of different modalities, learning indistinguishable representations or explicit modality transfer. The first two approaches suffer from the loss of discriminant information while removing the modality-specific variations. The third one heavily relies on the successful modality transfer, could face catastrophic performance drop when explicit modality transfers are not possible or difficult. To tackle this problem, we proposed a compact encoder-decoder neural module (cmUNet) to learn modality-agnostic representations while retaining identity-related information. This is achieved through cross-modality transformation and in-modality reconstruction, enhanced by an adversarial/perceptual loss which encourages indistinguishability of representations in the original sample space. For cross-modality matching, we propose MarrNet where cmUNet is connected to a standard feature extraction network which takes as inputs the modality-agnostic representations and outputs similarity scores for matching. We validated our method on five challenging tasks, namely Raman-infrared spectrum matching, cross-modality person re-identification and heterogeneous (photo-sketch, visible-near infrared and visible-thermal) face recognition, where MarrNet showed superior performance compared to state-of-the-art methods. Furthermore, it is observed that a cross-modality matching method could be biased to extract discriminant information from partial or even wrong regions, due to incompetence of dealing with modality gaps, which subsequently leads to poor generalization. We show that robustness to occlusions can be an indicator of whether a method can well bridge the modality gap. This, to our knowledge, has been largely neglected in the previous works. Our experiments demonstrated that MarrNet exhibited excellent robustness against disguises and occlusions, and outperformed existing methods with a large margin (>10%). The proposed cmUNet is a meta-approach and can be used as a building block for various applications.

Pseudo Label Association and Prototype-Based Invariant Learning for Semi-Supervised NIR-VIS Face Recognition.

Remarkable success of the existing Near-InfraRed and VISible (NIR-VIS) approaches owes to sufficient labeled training data. However, collecting and tagging data from different domains is a time-consuming and expensive task. In this paper, we tackle the NIR-VIS face recognition problem in a semi-supervised manner, termed as semi-supervised NIR-VIS Heterogeneous Face Recognition (NIR-VIS-sHFR). To cope with this problem, we propose a novel pseudo Label association and Prototype-based invariant Learning (LPL), consisting of three key components, i.e., Cross-domain pseudo Label Association (CLA), Intra-domain Compact Representation learning (ICR), and Prototype-based Inter-domain Invariant learning (PII). Firstly, the CLA iteratively builds inter-domain association graphs for pseudo-label association, subsequently facilitating cross-domain model development based on the generated pseudo-labels. Furthermore, the ICR is proposed to achieve the separation of in-domain features from different clusters and the aggregation of features from the same cluster, by performing cluster adaptation learning with prototype-based initialization. Finally, with the cross-domain pseudo-label training data produced by CLA, the PII explores potential domain-invariant and identity-related features, which employs cross-domain prototypes with identity-associated momentum updating to effectively guide inter-domain instances learning. The semi-supervised LPL method achieves comparable performance to recent supervised learning methods on multiple challenging NIR-VIS datasets, which demonstrates that the LPL is capable of learning robust cross-domain representations even without identity label information.

Pose aligned modality-invariant feature learning for NIR–VIS heterogeneous face recognition

Both pose disparities and modality differences with various modalities are significant difficulties that have an impact on recognition accuracy in heterogeneous face recognition. In this paper, we propose the pose aligned modality-invariant feature learning (PAMFL) method for NIR–VIS face recognition. This method disentangles the processing of the face pose and modality into independent stages. In the first phase, we construct the face pose alignment module (PAM). The built StyleGAN2-based generator incorporates pose estimation and feature mapping structures to alter the face shape in accordance with pose yaw angle instructions, eliminating the face pose misalignment. In the second phase, we build the modality-invariant feature learning module (MFLM). Modality-specific feature representations are learned using a pseudo-Siamese network in the shallow layer of the network, while modality-invariant feature representations are learned using a parameter sharing layer embedded in the deeper layer of the network. This module preserves all modality-invariant features while minimizing cross-modality variation. Finally, comparative experiments on BUAA VisNir, CASIA NIR–VIS 2.0 and Oulu CASIA NIR–VIS datasets validate that the proposed PAMFL shows advanced performance in overcoming face pose misalignment and improving heterogeneous face recognition accuracy.

Syncretic Space Learning Network for NIR-VIS Face Recognition

To overcome the technical bottleneck of face recognition in low-light scenarios, Near-InfraRed and VISible (NIR-VIS) heterogeneous face recognition is proposed for matching well-lit VIS faces with poorly lit NIR faces. Current cross-modal synthesis methods visually convert the NIR modality to the VIS modality and then perform face matching in the VIS modality. However, using a heavyweight GAN network on unpaired NIR-VIS faces may lead to high synthesis difficulty, low inference efficiency, and other problems. To alleviate the above problems, we simultaneously synthesize NIR and VIS images into modality-independent syncretic images and propose a novel syncretic space learning (SSL) model to eliminate the modal gap. First, Syncretic Modality Generator (SMG) synthesizes NIR and VIS images into syncretic images using channel-level convolution with a shallow CNN. In particular, the discriminative structural information is well preserved and the face quality can be further improved with small modal variations in a self-supervised learning manner. Second, Modality-adversarial Syncretic space Learning (MSL) projects NIR and VIS images into the syncretic space by a syncretic-modality adversarial learning strategy with syncretic pattern guided objective, so the modal gap of NIR-VIS faces can be effectively reduced. Finally, the Syncretic Distribution Consistency (SDC) constructed by NIR-syncretic, syncretic-syncretic, and VIS-syncretic consistency can enhance the intra-class compactness and learn discriminative representations. Extensive experiments on three challenging datasets demonstrate the effectiveness of the SSL method.

One-way multimodal image-to-image translation for heterogeneous face recognition

Cross-domain face matching is known as heterogeneous face recognition (HFR), which is more valuable than traditional face recognition and plays a key role in public safety. It remains a challenging problem due to insufficient heterogeneous data and large domain discrepancies. Recent work has focused on learning common features from different visual domains and synthesizing face images from other domains into the visible domain to reduce domain discrepancy. The former performs domain transformation during face recognition, which requires additional runtime. The latter has poorer performance limited by insufficient heterogeneous data. We propose a one-way multimodal image-to-image translation (OMIT) to tackle the problem of insufficient heterogeneous data. Specifically, we learn the generation of heterogeneous data from small-scale paired heterogeneous training data, and an identity-preserving loss is imposed on the generated heterogeneous images to ensure their identity consistency. Then, we transform large-scale visible data of abundant identity information into paired heterogeneous data, which addresses the lack of identity diversity due to small-scale heterogeneous data. The generated pairwise heterogeneous data can be directly used to train the HFR network to improve recognition accuracy and precision. Finally, our method exhibits generated quality improvements on BUAA, CASIA NIR–VIS 2.0, Oulu NIR–VIS, and Tufts face database. Heterogeneous datasets generated by OMIT improve the performance of HFR networks.

Facial feature embedded CycleGAN for VIS–NIR translation

Visible and near-infrared (VIS–NIR) heterogeneous face recognition remains a challenging task due to distinctions between spectral components of two modalities and insufficiently pairwise VIS–NIR data. Inspired by the cycle-consistent generative adversarial network (CycleGAN), this paper proposes a facial feature embedded CycleGAN to translate between VIS and NIR face images, aiming to enable the distributions of translated (fake) images to be similar as those of true images. To learn the particular feature of NIR or VIS domain while preserving common facial representation between VIS and NIR domains, a facial feature extractor (FFE), tailored specifically for extracting effective feature from face images, is embedded in the generator of original CycleGAN. For implementing the FFE, we use the MobileFaceNet which is pre-trained on a VIS face database. The domain-invariant feature learning is enhanced by proposing a new pixel consistency loss. Additionally, we establish a new WHU VIS–NIR database including varies in face rotation and expressions to enrich the insufficient training data. Moreover, experiments on the well-known Oulu-CASIA NIR–VIS database and our WHU VIS–NIR database validate the potential benefit of the proposed FFE-based CycleGAN (FFE-CycleGAN). In particular, we achieve 96.5% accuracy on Oulu-CASIA and 98.9% accuracy on WHU VIS–NIR.

Prepended Domain Transformer: Heterogeneous Face Recognition Without Bells and Whistles

Heterogeneous Face Recognition (HFR) refers to matching face images captured in different domains, such as thermal to visible images (VIS), sketches to visible images, near-infrared to visible, and so on. This is particularly useful in matching visible spectrum images to images captured from other modalities. Though highly useful, HFR is challenging because of the domain gap between the source and target domain. Often, large-scale paired heterogeneous face image datasets are absent, preventing training models specifically for the heterogeneous task. In this work, we propose a surprisingly simple, yet, very effective method for matching face images across different sensing modalities. The core idea of the proposed approach is to add a novel neural network block called Prepended Domain Transformer (PDT) in front of a pre-trained face recognition (FR) model to address the domain gap. Retraining this new block with few paired samples in a contrastive learning setup was enough to achieve state-of-the-art performance in many HFR benchmarks. The PDT blocks can be retrained for several source-target combinations using the proposed general framework. The proposed approach is architecture agnostic, meaning they can be added to any pre-trained FR models. Further, the approach is modular and the new block can be trained with a minimal set of paired samples, making it much easier for practical deployment. The source code and protocols will be made available publicly.

Holistic Feature Reconstruction-Based 3-D Attention Mechanism for Cross-Spectral Periocular Recognition

Recognition of individuals in cross-spectral environments involves situations where probe and gallery images are captured in distinct wavelength ranges and hence differ significantly in terms of appearance and illumination. In case of heterogeneous periocular images, alleviating this wide appearance gap and learning to extract illumination-invariant features from such local regions become cumbersome, giving rise to a challenging research problem. In this work, we design a novel holistic feature reconstruction-based attention module (H-FRAM) to refine and generate discriminative convolutional features. In contrast to existing spatial and channel attention mechanisms that compute 2-D or 1-D attention weights respectively, H-FRAM calculates the importance of each feature location by performing multi-linear principal component analysis on full 3-D tensor space. In H-FRAM, we claim that the feature reconstruction error, computed in a holistic manner, plays a crucial role in determining the relevance of feature locations. This reconstruction error is found by projecting the input feature map onto a multi-dimensional eigenspace that captures most of the important variations. To our knowledge, this is the first work which explores subspace learning approach in the context of 3-D attention mechanism to capture discriminative feature information. We have created an in-house cross-spectral periocular dataset containing visible and near-infrared images from 200 classes. The images are captured in unconstrained acquisition setup involving unsupervised eye and head movements as well as accessory variations (face masks and eyeglasses). Extensive experiments and ablation studies show that the proposed network achieves state-of-the-art recognition performances on the existing and in-house datasets for heterogeneous and homogeneous periocular recognition as well as heterogeneous face recognition.

Disguised heterogeneous face recognition using deep neighborhood difference relational network

Heterogeneous face recognition (HFR) aims to match a probe face image to a gallery of images in a different modality, such as visible to thermal. The complexity of the problem is further increased by the presence of facial disguise, one of the emerging problems in application of face biometrics. While several works have been proposed for HFR, disguised HFR remains one of the unexplored challenges of face recognition. Most of the proposed methods apply constraints on the network output for cross-modality image matching, consequently ignoring the image relationships within the network. The deep features within a network contain additional information which can be exploited for improved performance. With this motivation, we propose a Deep Neighborhood Difference Relational (DNDR) network and Joint Discrimination Loss for disguised HFR. The network calculates the difference relationships between cross-modality images in deep feature space. It implicitly learns to ignore the real-to-disguise feature relationships while focusing on real-to-real facial features for HFR. The proposed method learns the relationships between images of the same identity rather than optimizing the network on a hand-crafted metric loss function. For an input image pair, the DNDR network outputs identity representative embeddings of each image and a match probability of the image pair. The embedding distance and the match probability are then fused for more robust classification accuracy. Extensive experiments on publicly available heterogeneous face databases show the effectiveness of the proposed method for visible to infrared, visible to thermal, and visible to sketch face recognition.

Orthogonal Modality Disentanglement and Representation Alignment Network for NIR-VIS Face Recognition

Near-infrared and visual (NIR-VIS) face matching, as the most typical task in Heterogeneous Face Recognition (HFR), has attracted increasing attention in recent years. However, due to the large within-class discrepancies, including domain differences and residual discrepancies (i.e., lighting, expressions, occlusion, blurry, pose, etc), this is still a difficult task. Conventional NIR-VIS FR methods only focus on reducing the modality gap between cross-domain images, while neglecting to eliminate the residual variations. To better solve the above problems, this paper proposes a novel Orthogonal Modality Disentanglement and Representation Alignment (OMDRA) approach, which consists of three key components, including Modality-Invariant (MI) loss, Orthogonal Modality Disentanglement (OMD) and Deep Representation Alignment (DRA). Firstly, the MI loss is designed to learn modality-invariant and identity-discriminative representation, by increasing between-class separability and within-class compactness between NIR and VIS heterogeneous data. Secondly, the high-level Hybrid Facial Feature (HFF) layer of the backbone network is projected into two subspaces: the modality-related and identity-related subspaces. The OMD is designed to decouple modal information via an adversarial process, and we further impose Orthogonal Representation Decorrelation (ORD) to the OMD to decrease the correlation between identity representations and domain representations, as well as enhancing their representation capabilities. Finally, the DRA aims to eliminate the residual variations by performing a high-level representation alignment between non-neutral face and neutral face, which can effectively guides the network to learn discriminative and residual-invariant face representation. The joint scheme enables the disentanglement of modality variations, elimination of residual discrepancies, and the purification of identity information. Extensive experiments on challenging cross-domain databases indicate that our OMDRA method is superior to the state-of-the-art methods.

Part-facial relational and modality-style attention networks for heterogeneous face recognition

The purpose of heterogeneous face recognition (HFR) is to match face images of the same person from different modalities. Most HFR methods bridge the cross-modality variations with feature alignment by global feature representation learning, but ignore the content information of local features and modality-style information of face image for each modality, which limits the performance for HFR. The content information of local features not only contains the invariance of modality face features, but also can improve the stability of global face features, i.e., local features such as eyes, nose and mouth are steady and invariant. With this motivation, we propose a cross-modality dual-constraint (CMDC) approach that includes the part-facial relational attention network (PRAN) and modality-style attention network (MSAN). First, PRAN is designed to estimate the intrinsic structural relationships of local content features on each modality. It can extract discriminative local face features by capturing correlations within the face space of individual modality, and strengthen representations by contextual relationships across modalities. Secondly, we design the MSAN to capture the modality-style information for each modality, and then reduce the inter-modality differences by minimizing the distance of two modality-style features. Thirdly, to alleviate cross-modality variances and enhance intra-class compactness and inter-class divisibility, we propose the cross-modality dual-constrained loss (DCLoss) in the CMDC approach, which adds a global constraint to each sample distribution in the embedding space. Meanwhile, on the basis of focusing on modality-style information, DCLoss emphasizes the significance of category information. Extensive experiments on four datasets demonstrate the superior performance of our approach over the existing state-of-the-art. The code is available at https://github.com/JianYu777/CMDC.

SUPREAR-NET: Supervised Resolution Enhancement and Recognition Network

Heterogeneous face recognition is a challenging problem where the probe and gallery images belong to different modalities such as, low and high resolution, visible and near-infrared spectrum. A Generative Adversarial Network (GAN) enables us to learn an image to image transformation model for enhancing the resolution of a face image. Such a model would be helpful in a heterogeneous face recognition scenario. However, unsupervised GAN based transformation methods in their native formulation might alter useful discriminative information in the transformed face images. This affects the performance of face recognition algorithms when applied on the transformed images. We propose a Supervised Resolution Enhancement and Recognition Network (SUPREAR-NET), which does not corrupt the useful class-specific information of the face image and transforms a low resolution probe image into a high resolution one, followed by effective matching with the gallery using a trained discriminative model. We show the results for cross-resolution face recognition on three datasets including the FaceSurv face dataset, containing poor quality low resolution videos captured at a standoff distance up to 10 meters from the camera. On the FaceSurv, NIST MEDS and CMU MultiPIE datasets, the proposed algorithm outperforms recent unsupervised and supervised GAN algorithms.

Heterogeneous Visible-Thermal and Visible-Infrared Face Recognition Using Cross-Modality Discriminator Network and Unit-Class Loss.

Heterogeneous face recognition (HFR) aims to match face images across different imaging domains such as visible-to-infrared and visible-to-thermal. Recently, the increasing utility of nonvisible imaging has increased the application prospects of HFR in areas such as biometrics, security, and surveillance. HFR is a challenging variate of face recognition due to the differences between different imaging domains. While the current research has proposed image preprocessing, feature extraction, or common subspace projection for HFR, the optimization of these multi-stage methods is a challenging task as each step needs to be optimized separately and the performance error accumulates over each stage. In this paper, we propose a unified end-to-end Cross-Modality Discriminator Network (CMDN) for HFR. The proposed network uses a Deep Relational Discriminator module to learn deep feature relations for cross-domain face matching. Simultaneously, the CMDN is used to extract modality-independent embedding vectors for face images. The CMDN parameters are optimized using a novel Unit-Class Loss that shows higher stability and accuracy over other popular metric-learning loss functions. The experimental results on five popular HFR datasets demonstrate that the proposed method achieves significant improvement over the existing state-of-the-art methods.

Heterogeneous Face Recognition via Face Synthesis With Identity-Attribute Disentanglement

Heterogeneous Face Recognition (HFR) aims to match faces across different domains (e.g., visible to near-infrared images), which has been widely applied in authentication and forensics scenarios. However, HFR is a challenging problem because of the large cross-domain discrepancy, limited heterogeneous data pairs, and large variation of facial attributes. To address these challenges, we propose a new HFR method from the perspective of heterogeneous data augmentation, named Face Synthesis with Identity-Attribute Disentanglement (FSIAD). Firstly, the identity-attribute disentanglement (IAD) decouples face images into identity-related representations and identity-unrelated representations (called attributes), and then decreases the correlation between identities and attributes. Secondly, we devise a face synthesis module (FSM) to generate a large number of images with stochastic combinations of disentangled identities and attributes for enriching the attribute diversity of synthetic images. Both the original images and the synthetic ones are utilized to train the HFR network for tackling the challenges and improving the performance of HFR. Extensive experiments on five HFR databases validate that FSIAD obtains superior performance than previous HFR approaches. Particularly, FSIAD obtains 4.8% improvement over state of the art in terms of VR@FAR=0.01% on LAMP-HQ, the largest HFR database so far.

Towards Lightweight Pixel-Wise Hallucination for Heterogeneous Face Recognition.

Cross-spectral face hallucination is an intuitive way to mitigate the modality discrepancy in Heterogeneous Face Recognition (HFR). However, due to imaging differences, the hallucination inevitably suffers from a shape misalignment between paired heterogeneous images. Rather than building complicated architectures to circumvent the problem like previous works, we propose a simple yet effective method called Shape Alignment FacE (SAFE). Specifically, given an image, we align its shape to that of the paired one under the assistance of a 3D face model. The produced aligned pair enables us to train a lightweight generator that solely concentrates on spectrum translation with a pixel-wise supervision. However, since the 3D face model is powerless to attributes like the hair and glasses, there are still pixel discrepancies between the aligned pair. Given that, in the image space, we introduce a probabilistic pixel-wise loss that incorporates the discrepancies into a probabilistic distribution. Moreover, in order to alleviate the influence of the shape misalignment on spectrum translation, a spectrum optimal transport is performed in a shape-irrelevant latent space. Note that, in the final inference phase, except the lightweight generator, all other auxiliary modules are discarded. In addition to superior performance in qualitative synthesis and quantitative recognition, extensive experiments on 6 datasets demonstrate that our method also gains other two distinct advantages over existing state-of-the-art counterparts. The first is using a more lightweight generator. Compared with the state-of-the-art method, our method can achieve higher recognition results with 128x fewer parameters and 63x fewer FLOPs with only 4.58ms latency on a single TITAN-XP. The second is training on low-shot datasets such as Oulu-CASIA NIR-VIS that just contains 1,920 images from 20 identities. To the best of our knowledge, we are the first that can perform well on such a small-scale dataset. These advantages make our method more practical in the real world and further push boundaries of heterogeneous face recognition.

Memory-Modulated Transformer Network for Heterogeneous Face Recognition

Heterogeneous face recognition (HFR) aims at matching face images across different domains. It is challenging due to the severe domain discrepancies and overfitting caused by small training datasets. Some researchers apply a “recognition via generation” strategy and propose to solve the problem by translating images from a given domain into the visual domain. However, in many HFR tasks such as near-infrablack HFR, there is no paiblack data, which makes it an unsupervised generation. Pose variations, background differences, and many other factors present challenges. Moreover, the generated results lack diversity since many previous works regard this image translation as a “one-to-one” generation task. Considering the information deficiency in the input images, we propose to formulate this image translation process as a “one-to-many” generation problem. Specifically, we introduce reference images to guide the generation process. We propose a memory module to explore the prototypical style patterns of the reference domain. After self-supervised updating, the memory items are attentively aggregated to represent the style information. Moreover, to subtly fuse the contents of input images with the style of reference images, we propose a novel style transformer module. Specifically, we crop the encoded input and reference feature maps into patches, and use the style transformer to establish long-range dependencies between the input and reference patches. Thus, the style of every input patch is transferblack based on those of the most relevant reference patches. Extensive experiments on multiple datasets for various HFR tasks, including NIR-VIS, thermal-VIS, sketch-photo, and gray-RGB, are conducted. The robustness and effectiveness of the proposed MMTN are demonstrated both quantitatively and qualitatively.