Domain Generalizable Person Search Using Unreal Dataset (2404.00626v1)
Abstract: Collecting and labeling real datasets to train the person search networks not only requires a lot of time and effort, but also accompanies privacy issues. The weakly-supervised and unsupervised domain adaptation methods have been proposed to alleviate the labeling burden for target datasets, however, their generalization capability is limited. We introduce a novel person search method based on the domain generalization framework, that uses an automatically labeled unreal dataset only for training but is applicable to arbitrary unseen real datasets. To alleviate the domain gaps when transferring the knowledge from the unreal source dataset to the real target datasets, we estimate the fidelity of person instances which is then used to train the end-to-end network adaptively. Moreover, we devise a domain-invariant feature learning scheme to encourage the network to suppress the domain-related features. Experimental results demonstrate that the proposed method provides the competitive performance to existing person search methods even though it is applicable to arbitrary unseen datasets without any prior knowledge and re-training burdens.
- Metareg: Towards domain generalization using meta-regularization. In NIPS, 1006–1016.
- Pstr: End-to-end one-step person search with transformers. In CVPR, 9458–9467.
- Hierarchical online instance matching for person search. In AAAI, 10518–10525.
- Norm-aware embedding for efficient person search. In CVPR, 12615–12624.
- A simple framework for contrastive learning of visual representations. In PMLR, 1597–1607.
- Meta batch-instance normalization for generalizable person re-identification. In CVPR, 3425–3435.
- Learning to detect and track visible and occluded body joints in a virtual world. In ECCV, 430–446.
- Adversarially adaptive normalization for single domain generalization. In CVPR, 8208–8217.
- A Kernel Two-Sample Test. JMLR, 13(25): 723–773.
- Context-aware unsupervised clustering for person search. In BMVC.
- End-to-end trainable trident person search network using adaptive gradient propagation. In ICCV, 925–933.
- Weakly supervised person search with region siamese networks. In CVPR, 12006–12015.
- Dmrnet++: Learning discriminative features with decoupled networks and enriched pairs for one-step person search. TPAMI, 45(6): 7319–7337.
- Style normalization and restitution for generalizable person re-identification. In CVPR, 3143–3152.
- Oimnet++: Prototypical normalization and localization-aware learning for person search. In ECCV, 621–637.
- Learning to generalize: Meta-learning for domain generalization. In AAAI, 3490–3497.
- Domain adaptive person search. In ECCV, 302–318.
- Sequential end-to-end network for efficient person search. In AAAI, 2011–2019.
- Debiased batch normalization via gaussian process for generalizable person re-identification. In AAAI, 1729–1737.
- No-reference image quality assessment in the spatial domain. TIP, 21(12): 4695–4708.
- Unified deep supervised domain adaptation and generalization. In ICCV, 5715–5725.
- Uncertainty-guided model generalization to unseen domains. In CVPR, 6790–6800.
- Learning to learn single domain generalization. In CVPR, 12556–12565.
- Faster r-cnn: Towards real-time object detection with region proposal networks. In NIPS, 91–99.
- Batch normalization embeddings for deep domain generalization. PR, 135: 109115.
- Instance normalization: The missing ingredient for fast stylization. arXiv preprint arXiv:1607.08022.
- Generalizing to unseen domains via adversarial data augmentation. In NIPS, 5339–5349.
- Learning to diversify for single domain generalization. In ICCV, 834–843.
- Joint detection and identification feature learning for person search. In CVPR, 3415–3424.
- Exploring visual context for weakly supervised person search. In AAAI, 3027–3035.
- Cascade transformers for end-to-end person search. In CVPR, 7267–7276.
- Person re-identification in the wild. In CVPR, 1367–1376.