MARS: A Video Benchmark for Large-Scale Person Re-Identification

By Liang Zheng, Zhi Bie, Yifan Sun, Jingdong Wang, Chi Su, Shengjin Wang, Qi Tian. European Conference on Computer Vision, 2016

1,076 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/c0387e788a52f10bf35d4d50659cfa515d89fbec

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. Zhedong Zheng
4. Qi Tian
5. Shiliang Zhang
6. Wanli Ouyang
7. Mang Ye
8. Tao A. Xiang

In Computer Vision and Pattern Recognition

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

Citations in the corpus

• Deep Learning for Person Re-Identification: A Survey and Outlook (2020, methodology): “by images/videos captured by single-modality visible cameras in the closed-world setting [4], [5], [7], [33],”
• Unsupervised Person Re-identification (2017, methodology): “Another popular choice consists in training an identification network and extracting the intermediate output as a discriminative embedding [1], [9], [29], [30].”
• Horizontal Pyramid Matching for Person Re-identification (2018, methodology): “For the identification model [32, 39, 33, 27], it always yields superior accuracy compared with verification model.”
• Pose-Invariant Embedding for Deep Person Re-Identification (2017, methodology): “Trained alone using a standard CNN architecture [37, 41, 44], we show that PoseBox yields very decent re-ID accuracy.”
• Multi-Task Learning with Low Rank Attribute Embedding for Multi-Camera Person Re-Identification (2018, background): “There are also approaches dedicated to person reidentification in large camera networks involving more than two cameras [35], [36], [37], [38], [39], [40], [41].”
• Knowledge Distillation for End-to-End Person Search (2019, neutral): “More recent re-identification approaches are based on CNNs [1, 31, 62, 64] and mainly concerned with the estimation of a ID-feature embedding spaces, either via Siamese networks and contrastive losses [1, 9, 11, 15, 31, 36, 50, 56, 62], or with ID-classification and cross-entropy losses [54, 67].”
• Enhancing Person Re-identification in a Self-Trained Subspace (2017, background): “Person re-identification (re-ID) [Zheng et al. 2016a; Zheng et al. 2015a; Zheng et al. 2016b] aims to recognize an individual across spatially disjoint cameras.”
• Deep Cosine Metric Learning for Person Re-identification (2018, methodology): “Influence of pretraining on overall performance is studied in [35].”
• Cross-View Asymmetric Metric Learning for Unsupervised Person Re-Identification (2017, methodology): “In particular, we combined two existing RE-ID datasets [37, 36] to obtain a larger one which contains over 230,000 samples.”
• A Survey of Deep Active Learning (2020, methodology): “[4, 230] PRID [93], MARS [256], BDD100K [246], DukeMTMC-VideoReID [237], CityPersons [249], Caltech Pedestrian[52] Person Re-id”
• Learning Affective Features With a Hybrid Deep Model for Audio–Visual Emotion Recognition (2018, methodology): “Similar resize operation is also used in a previous work [58].”
• Pose Transferrable Person Re-identification (2018, methodology): “The source dataset we select in this work is the MARS dataset [45], which is a large video-based ReID dataset and has rich pose variations.”
• Quality Aware Network for Set to Set Recognition (2017, background): “In person re-identification, [20, 37, 41] use features generated by deep convolutional network and obtain state-of-the-art performance.”
• StrongSORT: Make DeepSORT Great Again (2022, methodology): “In the appearance branch, given detections in each frame, the deep appearance descriptor (a simple CNN), which is pretrained on the person reidentification dataset MARS [75], is applied to extract their appearance features.”
• Re-ranking Person Re-identification with k-Reciprocal Encoding (2017, methodology): “MARS [49] is the largest video-based re-ID benchmark dataset to date, containing 1,261 identities and around 20,000 video sequences.”
• In Defense of the Triplet Loss for Person Re-Identification (2017, methodology): “This means we directly optimize the network for the final task, which renders an additional metric learning step obsolete.”