3DPeS: 3D people dataset for surveillance and forensics

By Davide Baltieri, Roberto Vezzani, Rita Cucchiara. J-HGBU '11, 2011

311 citations recorded by the source corpus.

The interest of the research community in creating reference datasets for performance analysis is always very high. Although new datasets, collecting large amounts of video footage are spreading in surveillance and forensics, few bench-marks with annotation data are available for testing specific tasks and especially for 3D/multi-view analysis. In this paper we present 3DPeS, a new dataset for 3D/multi- view surveillance and forensic applications. This has been designed for discussing and evaluating research results in people re-identification and other related activities (people detection, people segmentation and people tracking). The new assessed version of the dataset contains hundreds of video sequences of 200 people taken from a multi-camera distributed surveillance system over several days, with different light conditions; each person is detected multiple times and from different points of view. In surveillance scenarios, the dataset can be exploited to evaluate people reacquisition, 3D body models and people activity reconstruction algorithms. In forensics it can be adopted too, by relaxing some constraints (e.g. real time) and neglecting some information (e.g. calibration). Some results on this new dataset are presented using state of the art methods for people re-identification as a benchmark for future comparisons.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/c92cb7a1e9a1d326c0af3161ae364ee669ed98af

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. Liang Zheng
2. Yi Yang
3. Qi Tian
4. Shaogang Gong
5. Shiliang Zhang
6. Xiaogang Wang
7. Xiatian Zhu
8. Wanli Ouyang

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Liang Zheng
3. Xiaogang Wang
4. Qi Tian
5. Dacheng Tao
6. Luc Van Gool
7. Wanli Ouyang
8. Chunhua Shen

Citations in the corpus

• Visual Tracking: an Experimental Survey (2013, unknown): “The recent 3DPeS dataset for people re-identification contains videos with more than 200 people walking as recorded from eight different cameras in very long video sequences [28], while the commonly used PETS-series [29] contains many videos divided by problem statement and surveillance application.”
• MARS: A Video Benchmark for Large-Scale Person Re-Identification (2016, methodology): “3DPES [3] and ETH [32] are of similar scales.”
• Saliency Weighted Features for Person Re-identification (2014, methodology): “We evaluated our approach on three publicly available benchmark datasets: the VIPeR dataset [17], the 3DPeS dataset [5] and the CHUK02 dataset [29].”
• Learning to rank in person re-identification with metric ensembles (2015, methodology): “The 3DPeS data set is designed mainly for people tracking and person re-identification [1].”
• Person Re-identification: Past, Present and Future (2016, methodology): “Several multi-shot re-ID datasets exist, e.g., ETH [122], 3DPES [123], PRID-2011 [86], iLIDS-VID [105], and MARS [21].”
• Attention-Aware Compositional Network for Person Re-identification (2018, methodology): “Following Spindle [49], we merge the training set of Market-1501 [55], CUHK01 [24], CUHK02 [23], CUHK03 [25], PSDB [43], Shinpuhkan [21], PRID [20], VIPeR [18], 3DPeS [5] and i-LIDS [56] for training, and then test on SenseReID.”
• Human Semantic Parsing for Person Re-identification (2018, methodology): “In addition to these datasets that were used in evaluation, we utilize 3DPeS [4], CUHK01 [23], CUHK02 [22], PRID [17], PSDB [41], Shinpuhkan [20] and VIPeR [15] datasets to augment our training data.”
• Learning Deep Feature Representations with Domain Guided Dropout for Person Re-identification (2016, background): “CUHK03 [23] 1467 21012 5252 100 100 CUHK01 [21] 971 1552 388 485 485 PRID [15] 385 2997 749 100 649 VIPeR [13] 632 506 126 316 316 3DPeS [5] 193 420 104 96 96 i-LIDS [50] 119 194 48 60 60 Shinpuhkan [18] 24 18004 4500”
• People reidentification in surveillance and forensics (2013, methodology): “The cylindrical model-based panoramic appearance map proposed by Gandhi and Trivedi [2006] and the recently proposed SARC3D body model [Baltieri et al. 2011c] allow a fully view-invariant appearance description.”
• A survey of approaches and trends in person re-identification (2014, unknown): “The 3DPes dataset [105] further extends the Sarc3D dataset by including 600 videos of 200 people taken from 8 static and calibrated cameras.”
• 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 [33], [34], [35], [36].”
• Spindle Net: Person Re-identification with Human Body Region Guided Feature Decomposition and Fusion (2017, methodology): “To evaluate the performance of the proposed Spindle Net, seven existing standard person ReID datasets are evaluated in our experiments, including CUHK03 [13], CUHK01 [12], PRID [10], VIPeR [9], 3DPeS [3], i-LIDS [34], and Market-1501 [33].”
• Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-identification (2018, methodology): “We first pretrain the ResNet-50 model on image-based person re-identification datasets, including CUHK01 [23], CUHK03 [24], 3DPeS [3], VIPeR [12], DukeMTMC-reID [51] and CUHK-SYSU [44].”
• Local Fisher Discriminant Analysis for Pedestrian Re-identification (2013, methodology): “The evaluation of the proposed method is carried out on VIPER [13], 3DPES [3] and CAVIAR [9] datasets.”
• Person Re-Identification Using Kernel-Based Metric Learning Methods (2014, methodology): “– We provide a comprehensive performance evaluation using four sets of features, three kernels (linear, χ(2) and RBF-χ(2)) and four challenging re-ID datasets: VIPeR [14], CAVIAR [8], 3DPeS [4] and iLIDS [30].”
• A Comparative Analysis of Object Detection Metrics with a Companion Open-Source Toolkit (2021, background): “Later on, advancements in object detection techniques led to their increasing adoption in areas that included the manufacturing industry with optical inspections [13], video surveillance [14], forensics [15,16], medical image analysis [17–19], autonomous vehicles [20], and traffic monitoring [21].”
• Enhancing Person Re-identification in a Self-Trained Subspace (2017, methodology): “3DPeS is a set of selected snapshots of the original video dataset [61], containing 192 people and 1011 images.”
• Video-Based Human Behavior Understanding: A Survey (2013, methodology): “These datasets include the 3-D People Surveillance Dataset (3DPES) [138], the MIT trajectory dataset [139], and the Edinburgh Informatics Forum Pedestrian Database (EIFPD) [140].”