Graph Correspondence Transfer for Person Re-Identification

Authors: Qin Zhou, Heng Fan, Shibao Zheng, Hang Su, Xinzhe Li, Shuang Wu, Haibin Ling

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on five benchmarks including VIPe R, Road, PRID450S, 3DPES and CUHK01 evidence the superior performance of GCT model over other state-of-the-art methods.
Researcher Affiliation Collaboration 1Shanghai Key Laboratory of Digital Media Processing and Transmission, Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai 200240, China 2Computer Science and Engineering, South China University of Technology, Guangzhou 510006, China 3Department of Computer & Information Sciences, Temple University, Philadelphia 19122, USA 4Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China 5You Tu Lad, Tencent, Shanghai 200233, China
Pseudocode No The paper describes the methods using text and mathematical equations, and includes flowcharts, but does not contain explicit pseudocode or an algorithm block.
Open Source Code No All the parameters will be available in the source code to be released for accessible reproducible research.
Open Datasets Yes We conduct extensive experiments on three challenging single-shot datasets (VIPe R, Road and PRID450S), and two multi-shot datasets (3DPES and CUHK01). The paper also provides citations for each dataset: VIPe R (Gray, Brennan, and Tao 2007), Road (Shen et al. 2015), PRID450S (Roth et al. 2014), 3DPES (Baltieri, Vezzani, and Cucchiara 2011), CUHK01 (Li, Zhao, and Wang 2012).
Dataset Splits Yes We adopt the common half-training and half-testing setting (K ostinger et al. 2012), and randomly split the dataset into two equal subsets. The training/testing sets are further divided into the probe and gallery sets according to their view information. On all the datasets, both the training/testing set partition and probe/gallery set partition are performed 10 times and average performance is recorded.
Hardware Specification Yes The proposed algorithm is implemented in Matlab on an Intel(R) Core(TM) i7-5820K CPU of 3.30GHz. We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan X Pascal GPU used for this research.
Software Dependencies No The paper states "The proposed algorithm is implemented in Matlab" but does not specify a version number for Matlab or other software dependencies with version numbers.
Experiment Setup Yes The λ in Eq.(6) is set to 2. The number of trees in the random forest model is 500. The best R for VIPe R, Road, PRID450S, 3DPES and CUHK01 datasets are 20, 5, 10, 20 and 20 respectively. The size of local patch is 32 × 24.