Visible Thermal Person Re-Identification via Dual-Constrained Top-Ranking
Authors: Mang Ye, Zheng Wang, Xiangyuan Lan, Pong C. Yuen
IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on two datasets demonstrate the superior performance compared to the state-of-the-arts. |
| Researcher Affiliation | Academia | 1 Department of Computer Science, Hong Kong Baptist University, Hong Kong 2 National Institute of Informatics, Japan |
| Pseudocode | No | The paper describes the proposed method in detail in Section 3, including network architecture and loss functions, but it does not include any formal pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code is available on the first author s website |
| Open Datasets | Yes | Two publicly available Reg DB dataset [Nguyen et al., 2017] and SYSU-MM01 [Wu et al., 2017b] are adopted for evaluation. |
| Dataset Splits | Yes | We follow the evaluation protocol in [Ye et al., 2018], where the dataset is randomly split into two halves, one for training and one for testing. For testing, the images from one modality were used as the gallery set while the ones from the other modality as the probe set. The procedure is repeated for 10 trials to achieve statistically stable results. SYSU-MM01 [Wu et al., 2017b] is a large-scale dataset... The training set contains 395 persons, with 22258 visible images and 11909 thermal images. The testing set contains 96 persons, with 3803 thermal images for query and 301 randomly selected visible images as gallery set. |
| Hardware Specification | No | The paper does not explicitly mention any specific hardware used for running experiments (e.g., GPU models, CPU types, memory). |
| Software Dependencies | No | The paper mentions using "Tensorflow" for implementation but does not specify its version or any other software dependencies with version numbers. |
| Experiment Setup | Yes | The size of the embedding fully connected layer is set as 1024 and the batch size is set as 64 for both datasets. Dropout rate is set as 0.5. Random cropping is utilized for data argumentation, where images are firstly resized to 256 256, and then a random cropped 227 227 image is fed into the network. We set the trade-off parameters as λ1 = 0.1 and λ2 = 1. Momentum optimizer is utilized for optimization, and the momentum is set to 0.9. The predefined cross-modality margin ρ1 is set to 0.5 while the intramodality margin ρ2 is set to 0.1. The initial learning rate is set as 0.001. The training step for Reg DB dataset is 5000 and SYSU-MM01 dataset is 50000. |