HSME: Hypersphere Manifold Embedding for Visible Thermal Person Re-Identification
Authors: Yi Hao, Nannan Wang, Jie Li, Xinbo Gao8385-8392
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct experiments on two cross-modality person re-identification datasets. Experimental results demonstrate that our method outperforms the state-of-the-art methods on two datasets. |
| Researcher Affiliation | Academia | Yi Hao, Nannan Wang, Jie Li, Xinbo Gao State Key Laboratory of Integrated Services Networks, School of Electronic Engineering, Xidian University, Xi an 710071, China State Key Laboratory of Integrated Services Networks, School of Telecommunications Engineering, Xidian University, Xi an 710071, China Corresponding author: Nannan Wang (nnwang@xidian.edu.cn) |
| Pseudocode | Yes | Algorithm 1 Training Feature Decorrelated HSMEnet |
| Open Source Code | No | The paper does not contain any explicit statement about providing open-source code for the described methodology, nor does it provide a link to a code repository. |
| Open Datasets | Yes | Two public cross-modality person reidentificaition datasets are adopted to evaluate our algorithm. (Nguyen et al. 2017) provided a visible-thermal dataset Reg DB... SYSU-MM01(Wu et al. 2017a) include RGB and infrared(IR) images... |
| Dataset Splits | Yes | We follow the evaluation protocol in (Ye et al. 2018a), where the dataset is randomly split into training dataset and testing dataset. We repeat this split procedure for 10 trials to achieve stable results. The training data and testing data is already splited by (Wu et al. 2017a). The training set contains 395 persons, and the testing set contains 96 persons. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory specifications) used for running the experiments. |
| Software Dependencies | No | The paper mentions using AlexNet and Momentum optimizers, but does not provide specific version numbers for these or any other software dependencies. |
| Experiment Setup | Yes | We use Alex Net as backbone net for both visible and thermal streams. The size of fully connected layer in shared layers is set as 1024 and the number of sample pairs in a mini-batch is set as 64 for both datasets. Dropout rate is set as 0.5. We also use random cropping for data augmentation. We set the scale factor of Sphere Loss as 5. Two Momentum optimizers are utilized for both visible and thermal streams. The artificial margin ρ is set to 0.5. The training step of stage1 is equal to stage2, which is 1000 for Reg DB and 10000 for SYSU-MM01 dataset. |