Online Joint Multi-Metric Adaptation From Frequent Sharing-Subset Mining for Person Re-Identification

Zhou, Jiahuan; Su, Bing; Wu, Ying

doi:10.1109/cvpr42600.2020.00298

Cited by 40 publications

(13 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Table I, we compare with 21 representative methods of weight pretraining and finetuning, which is also the current best models on Market-1501, including MSCAN [63], DF [64], SSM [65], SVDNet [66], GAN [67], PDF [68], TriNet [69], TriNet + Era. + Re-ranking [1], PCB [83], Omin [70], JointDG [71], IANet [72], CASN+PCB [73], CAMA [74], MHN-6 [75], AANet [76], P 2 -Net [77], PGFA [78], ISP [79], CBN [80], SNR [81], and M 3 +ResNet50 [82]. All settings of the above methods are consistent with the common training settings.…”

Section: A Comparison With Weight Pretraining and Finetuning In Perso...mentioning

confidence: 60%

Joint Learning of Neural Transfer and Architecture Adaptation for Image Recognition

Wang

Lin

Chen

et al. 2022

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

Current state-of-the-art visual recognition systems usually rely on the following pipeline: (a) pretraining a neural network on a large-scale dataset (e.g., ImageNet) and (b) finetuning the network weights on a smaller, task-specific dataset. Such a pipeline assumes the sole weight adaptation is able to transfer the network capability from one domain to another domain, based on a strong assumption that a fixed architecture is appropriate for all domains. However, each domain with a distinct recognition target may need different levels/paths of feature hierarchy, where some neurons may become redundant, and some others are re-activated to form new network structures. In this work, we prove that dynamically adapting network architectures tailored for each domain task along with weight finetuning benefits in both efficiency and effectiveness, compared to the existing image recognition pipeline that only tunes the weights regardless of the architecture. Our method can be easily generalized to an unsupervised paradigm by replacing supernet training with self-supervised learning in the source domain tasks and performing linear evaluation in the downstream tasks. This further improves the search efficiency of our method. Moreover, we also provide principled and empirical analysis to explain why our approach works by investigating the ineffectiveness of existing neural architecture search. We find that preserving the joint distribution of the network architecture and weights is of importance. This analysis not only benefits image recognition but also provides insights for crafting neural networks. Experiments on five representative image recognition tasks such as person re-identification, age estimation, gender recognition, image classification, and unsupervised domain adaptation demonstrate the effectiveness of our method.

show abstract

Section: A Comparison With Weight Pretraining and Finetuning In Perso...mentioning

confidence: 60%

Joint Learning of Neural Transfer and Architecture Adaptation for Image Recognition

Wang

Lin

Chen

et al. 2022

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

show abstract

“…For the fourth query image, we can see that the current dataset contains a large number of similar images of pedestrians (ranks 1-10). Even with human eyes, it is impossible to distinguish whether these images belong to the same person, but our method can still obtain relatively [53] ResNet152 85.9 73.3 PCB+RPP (ECCV18) [38] ResNet50 83.3 69.2 DuATM (CVPR18) [37] DenseNet121 81.8 64.6 PSE+ECN(CVPR18) [36] ResNet50 84.5 75.7 AANet (CVPR19) [17] ResNet152 87.7 74.3 DCDS(ICCV19) [46] ResNet101 87.6 75.5 CASN(CVPR19) [55] ResNet50 87.7 73.7 HPM (AAAI19) [66] ResNet50 86.6 74.3 MHN-PCB(ICCV19) [45] ResNet50 89.1 77.2 OSNet (ICCV19) [48] OSNET 88.6 73.5 MGN(ACMMM18) [42] ResNet50 88.7 78.4 ABDNet (ICCV19) [56] ResNet50 89.0 78.6 GCP(AAAI20) [47] ResNet50 89.7 78.6 SAN(AAAI20) [49] ResNet50 87.9 75.5 3DTANet (TCSVT20) [50] -89.9 78.4 M3+ ResNet50(CVPR20) [54] ResNet50 84.7 68.5 M3+DenseNet121(CVPR20) [54] DenseNet121 84.9 68.0 HOReID (CVPR20) [51] ResNet50 86.9 75.6 Ours ResNet50 90.2 80.2…”

Section: Methodsmentioning

confidence: 99%

Hybrid-attention guided network with multiple resolution features for person re-identification

Guo-qing

Yang

Zheng

et al. 2021

Information Sciences

View full text Add to dashboard Cite

“…Evaluation Metrics For the ReID task, we follow the same evaluation protocols which are used in previous works [27,52,55]. Mean Average Precision (mAP) and the cumulative matching characteristics at Rank1 (CMC@1) are employed to evaluate the performance of PGVR.…”

Section: Datasets and Evaluation Metricsmentioning

confidence: 99%

Pseudo Graph Convolutional Network for Vehicle ReID

Qian

Peng

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Image-based Vehicle ReID methods have suffered from limited information caused by viewpoints, illumination, and occlusion as they usually use a single image as input. Graph convolutional methods (GCN) can alleviate the aforementioned problem by aggregating neighbor samples' information to enhance the feature representation. However, it's uneconomical and computational for the inference processes of GCN-based methods since they need to iterate over all samples for searching the neighbor nodes. In this paper, we propose the first Pseudo-GCN Vehicle ReID method (PGVR) which enables a CNN-based module to performs competitively to GCN-based methods and has a faster and lightweight inference process. To enable the Pseudo-GCN mechanism, a two-branch network and a graph-based knowledge distillation are proposed. The two-branch network consists of a CNN-based student branch and a GCN-based teacher branch. The GCN-based teacher branch adopts a ReID-based GCN to learn the topological optimization ability under the supervision of ReID tasks during training time. Moreover, the graph-based knowledge distillation explicitly transfers the topological optimization ability from the teacher branch to the student branch which acknowledges all nodes. We evaluate our proposed method PGVR on three mainstream Vehicle ReID benchmarks and demonstrate that PGVR achieves state-of-the-art performance. CCS CONCEPTS• Computing methodologies → Computer vision; Neural networks; • Information systems → Retrieval models and ranking.

show abstract

Online Joint Multi-Metric Adaptation From Frequent Sharing-Subset Mining for Person Re-Identification

Cited by 40 publications

References 27 publications

Joint Learning of Neural Transfer and Architecture Adaptation for Image Recognition

Joint Learning of Neural Transfer and Architecture Adaptation for Image Recognition

Hybrid-attention guided network with multiple resolution features for person re-identification

Pseudo Graph Convolutional Network for Vehicle ReID

Contact Info

Product

Resources

About