Reliability Exploration with Self-Ensemble Learning for Domain Adaptive Person Re-identification
Authors: Zongyi Li, Yuxuan Shi, Hefei Ling, Jiazhong Chen, Qian Wang, Fengfan Zhou1527-1535
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experiments demonstrate that our method can significantly surpass the state-of-the-art performance on the unsupervised domain adaptive person Re ID. Experiments show that our method achieves a tradeable effect and surpasses most state-of-the-art methods by large margins on multiple benchmarks of unsupervised domain adaptive Re-ID. Experiments Datasets and Evaluation Protocol We evaluate our method on three large-scale Re-ID datasets: Market-1501 (Zheng et al. 2015), Duke MTMC-Re ID (Ristani et al. 2016; Zheng, Zheng, and Yang 2017) and MSMT17 (Wei et al. 2018). Ablation Studies In this section, we perform ablation experiments on different components of our method to evaluate their effectiveness. |
| Researcher Affiliation | Academia | Zongyi Li, Yuxuan Shi*, Hefei Ling, Jiazhong Chen, Qian Wang, Fengfan Zhou Department of Computer Science and Technology, Huazhong University of Science and Technology {zongyili, shiyx, lhefei, jzchen, yqwq1996, fengfanzhou}@hust.edu.cn |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statement or link for open-source code availability. |
| Open Datasets | Yes | We evaluate our method on three large-scale Re-ID datasets: Market-1501 (Zheng et al. 2015), Duke MTMC-Re ID (Ristani et al. 2016; Zheng, Zheng, and Yang 2017) and MSMT17 (Wei et al. 2018). |
| Dataset Splits | Yes | Market-1501 (Zheng et al. 2015) consists of 1501 identities with 32,668 images... The training set contains 751 identities with 12,936 images. The test set includes 750 identities, where the query set contains 3,368 images and the gallery set contains 19,732 images. Duke MTMC-Re ID (Ristani et al. 2016)... this dataset contains 16,522 training images, 2,228 queries images, and 17,661 gallery images. MSMT17 (Wei et al. 2018)... The train set contains 1,041 identities and test set contains 3,060 identities. |
| Hardware Specification | No | The paper mentions using ResNet50 as a CNN architecture and discusses memory usage, but it does not specify any particular hardware (e.g., GPU models, CPU types, or server specifications) used for running the experiments. |
| Software Dependencies | No | The paper mentions the SGD optimizer but does not specify any software names with version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions) required to replicate the experiment. |
| Experiment Setup | Yes | The number of K is set to 3, and each image has three augmented samples in each training batch. The batch size for source domain and target domain are both set as 128 by sampling 32 identities and 4 image per identity. We utilize the DBSCAN clustering algorithm, and the Jaccard distance with k-reciprocal nearest neighbor is used as the distance metric. The eps in DBSACN is set to 0.6. SGD optimizer is adopted for model optimization. The initial learning rate is 0.00035, and is divided by 10 at the 40th and 60th epoch, in a total 70 epochs. The reliability threshold β is set to 0.6. the momentum factor α and m in Eq.(4) and Eq.(11) are set to 0.999 and 0.2. |