Multi-level Factorisation Net for Person Re-identification

By Xiaobin Chang, Timothy M. Hospedales, Tao A. Xiang. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018

526 citations recorded by the source corpus.

Key to effective person re-identification (Re-ID) is modelling discriminative and view-invariant factors of person appearance at both high and low semantic levels. Recently developed deep Re-ID models either learn a holistic single semantic level feature representation and/or require laborious human annotation of these factors as attributes. We propose Multi-Level Factorisation Net (MLFN), a novel network architecture that factorises the visual appearance of a person into latent discriminative factors at multiple semantic levels without manual annotation. MLFN is composed of multiple stacked blocks. Each block contains multiple factor modules to model latent factors at a specific level, and factor selection modules that dynamically select the factor modules to interpret the content of each input image. The outputs of the factor selection modules also provide a compact latent factor descriptor that is complementary to the conventional deeply learned features. MLFN achieves state-of-the-art results on three Re-ID datasets, as well as compelling results on the general object categorisation CIFAR-100 dataset.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/004acfec16c36649408c561faa102dd9de76f085

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. Luc Van Gool
2. Tao A. Xiang
3. Dacheng Tao
4. Xiaogang Wang
5. Chen Change Loy
6. Yu Qiao
7. Yi Yang
8. Ming-Hsuan Yang

Citations in the corpus

• Batch Coherence-Driven Network for Part-Aware Person Re-Identification (2020, unknown): “0 MLFN [64] 90.”
• Robust Person Re-Identification by Modelling Feature Uncertainty (2019, methodology): “Deep Person ReID Models Existing deep ReID models [2, 3, 7, 10, 18, 24, 27, 30, 33, 39] typically adopt an off-the-shelf CNN architecture for deep feature learning, with recent methods mostly using ResNet [6].”
• Omni-Scale Feature Learning for Person Re-Identification (2019, methodology): “Different to [2], the network in [30] is built on Inception [44, 45], where multiple streams were originally designed for low computational cost with handcrafted mixture of convolution and pooling layers.”
• Interaction-And-Aggregation Network for Person Re-Identification (2019, methodology): “[4] directly fused the feature maps across multiple layers to generate a single feature.”
• Learning Discriminative Features with Multiple Granularities for Person Re-Identification (2018, background): “Some recent deep Re-ID methods [4, 5, 22, 27, 29, 30, 36] have achieved breakthrough with high-level identification rates and mean average precision.”
• Dynamic Anchor Feature Selection for Single-Shot Object Detection (2019, background): “Object detection is a prerequisite for many downstream computer vision applications, such as person reidentification [2], autonomous driving [7], and action recognition [12].”
• ABD-Net: Attentive but Diverse Person Re-Identification (2019, unknown): “25 MLFN [59] (2018 CVPR) 90.”
• Hierarchical Clustering With Hard-Batch Triplet Loss for Person Re-Identification (2019, background): “With the development of convolutional neural networks (CNN), supervised re-ID [9, 11, 21, 25, 30, 2, 31, 23] has achieved excellent performance.”
• Bilinear Attention Networks for Person Retrieval (2019, unknown): “8 - MLFN [3] 74.”
• Domain Generalization: A Survey (2021, methodology): “A common practice is to combine several face anti-spoofing datasets for model training and do evaluation on an unseen dataset, e.g., using CASIA-MFSD [79], Oulu-NPU [80] and MSU-MFSD [81] as the sources and Idiap Replay-Attack [82] as the target.”
• Hyperbolic Image Embeddings (2019, methodology): “A number of methods utilize a simple classification approach for training [3, 33, 11, 45], and Euclidean distance is used in test time.”
• Memory-Based Neighbourhood Embedding for Visual Recognition (2019, unknown): “5 MLFN [1] CVPR’18 62.”
• Mode-adaptive neural networks for quadruped motion control (2018, background): “Recently its combination with deep learning architectures [Chang et al. 2018; Eigen et al. 2013; Shazeer et al. 2017] is showing promising results.”