Neighborhood Consensus Contrastive Learning for Backward-Compatible Representation
Authors: Shengsen Wu, Liang Chen, Yihang Lou, Yan Bai, Tao Bai, Minghua Deng, Ling-Yu Duan2722-2730
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our proposed method on two widely-used person Re ID datasets, i.e., Market-1501 (Zheng et al. 2015), MSMT17 (Wei et al. 2018), and one vehicle Re ID dataset Ve Ri-776 (Wang et al. 2017). We first implement several baselines, and then test the potential of our method by applying it to the case of multi-factor changes, including model changes and loss changes. We also conduct a multi-model test to evaluate the sequential compatibility. |
| Researcher Affiliation | Collaboration | 1The SECE of Shenzhen Graduate School, Peking University, Shenzhen, China 2National Engineering Research Center of Visual Technology, School of Computer Science, Peking University, Beijing, China 3School of Mathematical Sciences, Peking University, Beijing, China 4Intelligent Vision Dept, Huawei Technologies, Beijing, China 5Peng Cheng Laboratory, Shenzhen, China |
| Pseudocode | No | The paper does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not include an unambiguous statement about releasing the source code for its methodology or provide a direct link to a code repository. |
| Open Datasets | Yes | We evaluate our proposed method on two widely-used person Re ID datasets, i.e., Market-1501 (Zheng et al. 2015), MSMT17 (Wei et al. 2018), and one vehicle Re ID dataset Ve Ri-776 (Wang et al. 2017). |
| Dataset Splits | No | The paper discusses different training data subsets (e.g., '50% IDs subset' for old training data and '100% IDs complete set' for new training data, or '25% IDs subset' vs '75% IDs subset'), but it does not provide explicit overall train/validation/test dataset split percentages or sample counts for the datasets used. |
| Hardware Specification | No | The paper does not provide any specific details regarding the hardware (e.g., GPU models, CPU types, memory) used to run the experiments. |
| Software Dependencies | No | The paper mentions 'Fast Re ID', 'SBS', and 'bagtricks' frameworks but does not provide specific version numbers for these software components or any other libraries. |
| Experiment Setup | Yes | The size of B is 2048 and τ is 1.0. We set α and β around 0.01 so that L1 loss and L2 loss are on the same order of magnitude as Lnew loss. ˆU is set to log Knew / 2 as the credible sample selection threshold. |