Hierarchical Discriminative Learning for Visible Thermal Person Re-Identification

By Mang Ye, Xiangyuan Lan, Jiawei Li, Pong Chi Yuen. AAAI Conference on Artificial Intelligence, 2018

410 citations recorded by the source corpus.

Person re-identification is widely studied in visible spectrum, where all the person images are captured by visible cameras. However, visible cameras may not capture valid appearance information under poor illumination conditions, e.g, at night. In this case, thermal camera is superior since it is less dependent on the lighting by using infrared light to capture the human body. To this end, this paper investigates a cross-modal re-identification problem, namely visible-thermal person re-identification (VT-REID). Existing cross-modal matching methods mainly focus on modeling the cross-modality discrepancy, while VT-REID also suffers from cross-view variations caused by different camera views. Therefore, we propose a hierarchical cross-modality matching model by jointly optimizing the modality-specific and modality-shared metrics. The modality-specific metrics transform two heterogenous modalities into a consistent space that modality-shared metric can be subsequently learnt. Meanwhile, the modality-specific metric compacts features of the same person within each modality to handle the large intra-modality intra-person variations (e.g. viewpoints, pose). Additionally, an improved two-stream CNN network is presented to learn the multi-modality sharable feature representations. Identity loss and contrastive loss are integrated to enhance the discriminability and modality-invariance with partially shared layer parameters. Extensive experiments illustrate the effectiveness and robustness of the proposed method.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/cdb76fde739b102507140bd330e576bfb49166b7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Robust Anchor Embedding for Unsupervised Video Person re-IDentification in the Wild (2018, methodology): “Existing deep learning re-ID methods can be roughly categorized into three categories according to the learning objectives: triplet loss [5, 11, 48], contrastive/verification loss [6, 35, 45] and classification/identity loss [33, 56].”
Deep Learning for Person Re-Identification: A Survey and Outlook (2020, background): “A two stream network is introduced in [146], [190] to model the modality-sharable and -speciﬁc information, addressing the intra-and cross-modality variations simultaneously.”
Channel Augmented Joint Learning for Visible-Infrared Recognition (2021, background): “Visible-Infrared Person Re-IDentification (VI-ReID) is a cross-modality person recognition problem, which aims at matching daytime visible and night-time infrared images [2, 48, 16, 57].”
Hi-CMD: Hierarchical Cross-Modality Disentanglement for Visible-Infrared Person Re-Identification (2019, methodology): “We followed the RegDB evaluation protocol in [37, 39] and the SYSU-MM01 evaluation protocol Datasets RegDB [21] SYSU-MM01 [33] Methods R=1 R=10 mAP R=1 R=10 mAP”
RGB-Infrared Cross-Modality Person Re-Identification via Joint Pixel and Feature Alignment (2019, methodology): “To deal with the above issues, existing cross-modality re-id methods [32, 37, 38, 3, 9] mainly focus on bridging the gap between the RGB and IR images via feature alignment as shown in Figure 1.”
Cross-Modality Paired-Images Generation for RGB-Infrared Person Re-Identification (2020, methodology): “Ye et al. utilize a dual-path network with a bi-directional dual-constrained top-ranking loss (Ye et al. 2018a) and modality-speciﬁc and modality-shared metrics (Ye et al. 2018b).”
Learning to Reduce Dual-Level Discrepancy for Infrared-Visible Person Re-Identification (2019, methodology): “[22] introduced a two-step framework for feature learning and metric learning.”
Infrared-Visible Cross-Modal Person Re-Identification with an X Modality (2020, methodology): “For statistically stable results, the evaluation procedure is repeated for 10 trials with randomly sampled query and gallery sets, as (Ye et al. 2018a; Feng, Lai, and Xie 2019; Wang et al. 2019c).”
Bi-Directional Center-Constrained Top-Ranking for Visible Thermal Person Re-Identification (2020, methodology): “In particular, our proposed method also outperforms the two-stage learning method (TONE + HCML) [18] usually by a large margin for both rank-1 accuracy and mAP on two datasets.”
Cross-Modality Person Re-Identification With Shared-Specific Feature Transfer (2020, methodology): “In this subsection, we compare our proposed algorithm with the baselines as well as the state-of-the-art methods, including Zero-Padding [46], TONE [49], BDTR [50], cm-GAN [4], D 2 RL[44], MSR[7], D-HSME[13], IPVT[19], JSIA-ReID[57] and AlignGAN[42].”
Dynamic Dual-Attentive Aggregation Learning for Visible-Infrared Person Re-Identification (2020, methodology): “However, the modality difference of VI-ReID is much greater than that of face recognition due to the different camera environments and large visual appearances change, which limits the applicability of their methods to the VI-ReID [57,48].”