Person Re-identification by Multi-Channel Parts-Based CNN with Improved Triplet Loss Function

By De Cheng, Yihong Gong, Sanping Zhou, Jinjun Wang, Nanning Zheng. Computer Vision and Pattern Recognition, 2016

1,326 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/376aad5815f808c82e518956a70091bf828dbd25

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

1. Yi Yang
2. Liang Zheng
3. Qi Tian
4. Shiliang Zhang
5. Zhedong Zheng
6. Wanli Ouyang
7. Tao A. Xiang
8. Mang Ye

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Luc Van Gool
3. Dacheng Tao
4. Xiaogang Wang
5. Ming-Hsuan Yang
6. Liang Zheng
7. Bernt Schiele
8. Qi Tian

Citations in the corpus

• Infrared-Visible Cross-Modal Person Re-Identification with an X Modality (2020, background): “Person re-identification (ReID) aims at identifying target persons in a query set from a large-scale gallery set captured by non-overlapping camera views (Zheng, Yang, and Hauptmann 2016; Zhang et al. 2015; Cheng et al. 2016).”
• A Discriminatively Learned CNN Embedding for Person Reidentification (2016, background): “[7] split the convolutional map into four parts and fuse the part features with the global features.”
• Human Semantic Parsing for Person Re-identification (2018, background): “Others [12, 9, 11] have employed triplet loss which requires a tuple of anchor, positive and negative images where the training objective is to simultaneously pushing the positive image towards the anchor while pulling the negative image away from it.”
• Multi-scale Deep Learning Architectures for Person Re-identification (2017, methodology): “Similarly, [4] employed triplet loss to integrate multi-channel parts-based CNN models.”
• Spindle Net: Person Re-identification with Human Body Region Guided Feature Decomposition and Fusion (2017, methodology): “The proposed Spindle Net is compared with several state-of-the-art ReID approaches in recent two years, including the metric learning method WARCA-χ2 [11], the null space semi-supervised learning method NFST [32], the deep neural network PersonNet [28], the Siamese CNN with learnable gate S-CNN [25]”
• AlignedReID: Surpassing Human-Level Performance in Person Re-Identification (2017, background): “With the renaissance of deep learning, the convolutional neural network (CNN) has dominated this field [24, 32, 6, 54, 16, 24], by learning features in an end-to-end fashion through various metric learning losses such as contrastive loss [32], triplet loss [18], improved triplet loss [6], quadruplet”
• PifPaf: Composite Fields for Human Pose Estimation (2019, neutral): “Some prior work has used part-based or region-based models [45, 7, 43] that would profit from quality pose estimates.”
• See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-Based Person Re-identification (2017, neutral): “A large portion of existing works lies in the former category [7, 21, 28, 32, 33, 34] while only a few belong to the latter.”
• Image Aesthetic Assessment: An experimental survey (2016, background): “It is widely used in many recent vision studies [55], [70], [71], [72], aiming to bring data of the same class closer, while data of different classes further away.”
• Pedestrian Alignment Network for Large-scale Person Re-Identification (2017, methodology): “For methods that use patch matching, it is assumed that the matched local structures locate in the same horizontal stripe [Li et al., 2014, Yi et al., 2014, Zhao et al., 2013a, Zhao et al., 2014, Liao et al., 2015, Cheng et al., 2016] or square neighborhood [Ahmed et al., 2015].”
• Learning Deep Neural Networks for Vehicle Re-ID with Visual-spatio-Temporal Path Proposals (2017, background): “[3] trained CNN with triplet samples and minimized feature distances between the same person and maximize the distances between different people.”
• Pose-Driven Deep Convolutional Model for Person Re-identification (2017, methodology): “Apart from deep metric learning methods, some algorithms first learn image representations directly with the Triplet Loss or the Siamese Contrastive Loss, then utilize Euclidean distance for comparison [48, 5, 10, 46].”
• A Pose-Sensitive Embedding for Person Re-identification with Expanded Cross Neighborhood Re-ranking (2017, background): “We studied the impact of changing these on Market and Duke datasets and found that it is subtle in the range for t ∈ [2, 4] and q ∈ [4, 10], the performance drops between∼0.”
• Learning Deep Context-Aware Features over Body and Latent Parts for Person Re-identification (2017, background): “Some DCNN models [5, 31, 44] adopt rigid body parts to learn local pedestrian features.”
• Contrastive Adaptation Network for Unsupervised Domain Adaptation (2019, neutral): “face recognition [6], person re-identification [16], etc.”
• Beyond Triplet Loss: A Deep Quadruplet Network for Person Re-identification (2017, methodology): “The triplet loss is designed to keep xi closer to xj than xk, and widely used in many areas, such as image retrieval [33], face recognition [26] and person re-identification [7, 4].”