Uncertainty-Aware Multi-Shot Knowledge Distillation for Image-Based Object Re-Identification
Authors: Xin Jin, Cuiling Lan, Wenjun Zeng, Zhibo Chen11165-11172
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate the effectiveness of our approach on the popular vehicle re-id and person re-id datasets. In inference, the S-net alone significantly outperforms the baselines and achieves the state-of-the-art performance. We conduct extensive ablation studies and demonstrate the effectiveness of the framework and components on both person re-id and vehicle re-id datasets. |
| Researcher Affiliation | Collaboration | Xin Jin,1 Cuiling Lan,2 Wenjun Zeng,2 Zhibo Chen1 University of Science and Technology of China1 Microsoft Research Asia2 jinxustc@mail.ustc.edu.cn, {culan, wezeng}@microsoft.com, chenzhibo@ustc.edu.cn |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any concrete access to source code for the methodology described, nor does it include a specific repository link or an explicit code release statement. |
| Open Datasets | Yes | We conduct object re-id experiments on the most commonly-used person re-id dataset, CUHK03 (Li et al. 2014) (including the labeled/detected bounding box settings), and three vehicle re-id datasets of Ve Ri-776 (Liu et al. 2016), Vehicle ID (Liu, Tian, and others 2016) and the recent large-scale VERI-Wild (Lou et al. 2019). |
| Dataset Splits | No | The paper mentions using common practices for evaluation but does not specify exact split percentages or sample counts for training, validation, or test sets needed to reproduce the data partitioning. |
| Hardware Specification | Yes | All our models are implemented on Py Torch and trained on a single NVIDIA-P40 GPU. |
| Software Dependencies | No | The paper states, "All our models are implemented on Py Torch", but does not provide specific version numbers for PyTorch or any other software libraries, which is required for reproducible software dependencies. |
| Experiment Setup | Yes | We set K as 4 and add UA-KDLs at all the 5 stages by default. The batch size is set as 64. The input image resolution is set to 256 x 256 for vehicle re-id and 256 x 128 for person re-id, respectively. We use the commonly used data augmentation strategies of random cropping (Zhang et al. 2019), horizontal flipping, label smoothing regularization (Szegedy et al. 2016), and random erasing (Zhong et al. 2017) in both the baseline schemes and our schemes. We use Adam optimizer (Kingma and Ba 2014) for model optimization. |