Joint Discriminative and Generative Learning for Person Re-Identification

By Zhedong Zheng, Xiaodong Yang, Zhiding Yu, Liang Zheng, Yi Yang, Jan Kautz. Computer Vision and Pattern Recognition, 2019

846 citations recorded by the source corpus.

Person re-identification (re-id) remains challenging due to significant intra-class variations across different cameras. Recently, there has been a growing interest in using generative models to augment training data and enhance the invariance to input changes. The generative pipelines in existing methods, however, stay relatively separate from the discriminative re-id learning stages. Accordingly, re-id models are often trained in a straightforward manner on the generated data. In this paper, we seek to improve learned re-id embeddings by better leveraging the generated data. To this end, we propose a joint learning framework that couples re-id learning and data generation end-to-end. Our model involves a generative module that separately encodes each person into an appearance code and a structure code, and a discriminative module that shares the appearance encoder with the generative module. By switching the appearance or structure codes, the generative module is able to generate high-quality cross-id composed images, which are online fed back to the appearance encoder and used to improve the discriminative module. The proposed joint learning framework renders significant improvement over the baseline without using generated data, leading to the state-of-the-art performance on several benchmark datasets.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/ec721b4b280ce593428499d013bc01ca19dbcac3

How researchers see this paper

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

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Tao A. Xiang
3. Luc Van Gool
4. Ling Shao
5. Xiaogang Wang
6. Dacheng Tao
7. Ming-Hsuan Yang
8. Liang Zheng

Citations in the corpus

• Boundary-aware Backward-Compatible Representation via Adversarial Learning in Image Retrieval (2023, background): “Adversarial learning is widely used in various fields such as generative adversarial networks [4, 6, 7], person re-identity [33], and domain adaptation [2, 5].”
• Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification (2019, background): “The art supervised methods have made great achievement in person re-ID [14, 26, 39, 25], relying on rich-labeled data and the success of deep networks [18, 12, 8].”
• Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID (2020, methodology): “5 DG-Net [59] CVPR’19 52.”
• Cross-Modality Person Re-Identification With Shared-Specific Feature Transfer (2020, positive): “Recent disentangle based methods extend along this direction further by splitting each sample to identity-related and identitiy-unrelated features, obtaining purer representations without redundant cues [12, 53].”
• Batch Coherence-Driven Network for Part-Aware Person Re-Identification (2020, neutral): “3 DG-Net [76] 86.”
• RGB-Infrared Cross-Modality Person Re-Identification via Joint Pixel and Feature Alignment (2019, methodology): “Existing methods can be grouped into hand-crafted descriptors [22, 36, 19, 34], metric learning methods [43, 16, 20] and deep learning algorithms [42, 11, 28, 21, 44, 30, 15, 35, 33, 18, 17, 1, 31].”
• Cross-Modality Paired-Images Generation for RGB-Infrared Person Re-Identification (2020, background): “Zheng et al. (Zheng, Yang, and Hauptmann 2016) learn identity-discriminative features by fine-tuning a pre-trained CNN to minimize a classification loss.”
• Learning an Evolutionary Embedding via Massive Knowledge Distillation (2020, methodology): “Although Mancs (Wang et al. 2018), DSA-reID (Zhang et al. 2019) and DG-Net (Zheng et al. 2019) also obtain good performance, our EEL only utilizes a single network to learn embeddings under the supervision of Eq.”
• Deep Learning for Person Re-Identification: A Survey and Outlook (2020, methodology): “Specifically, we include PCB [68], MGN [179], Auto-ReID [144], ABD-Net [137], BagTricks [151], OSNet [143], DGNet [117], SCAL [83], MHN [101], P2Net [100], BDB [121], SONA [102], SFT [89], ConsAtt [87], DenseS [99], Pyramid [162], IANet [103].”
• Patch-based Knowledge Distillation for Lifelong Person Re-Identification (2022, background): “Person ReID has made remarkable progress under various settings such as supervised learning [18, 40, 57], unsupervised learning [9, 36] and unsupervised domain adaptation [12, 15, 66, 75, 76].”
• Hi-CMD: Hierarchical Cross-Modality Disentanglement for Visible-Infrared Person Re-Identification (2019, methodology): “While DG-Net decomposes each RGB image into appearance and structure codes, our Hi-CMD hierarchically disentangles ID-discriminative factors and ID-excluded factors including pose and illumination attributes from RGBIR images.”
• Counterfactual Attention Learning for Fine-Grained Visual Categorization and Re-identification (2021, neutral): “For MSMT17, we followed the settings in the test settings in [60] and [67].”
• Image-to-Image Translation: Methods and Applications (2021, methodology): “I2I methods can also contribute to other visual tasks, such as transfer learning for reinforcement learning [217], image registration [38], domain adaptation [29], [30], [31], person re-identification [218], [219], [220], [221], [222], image segmentation [7], [8], [9], facial geometry reconstruction”
• Omni-Scale Feature Learning for Person Re-Identification (2019, methodology): “As OSNet is orthogonal to some methods, such as the image generation based DGNet [77], they can be potentially combined to further boost the ReID performance.”