Vision Shared and Representation Isolated Network for Person Search
Authors: Yang Liu, Yingping Li, Chengyu Kong, Yuqiu Kong, Shenglan Liu, Feilong Wang
IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The proposed method outperforms the state-of-the-art methods on the mainstream datasets. |
| Researcher Affiliation | Academia | 1School of Innovation and Entrepreneurship, Dalian University of Technology 2Faculty of Electronic Information and Electrical Engineering, Dalian University of Technology |
| Pseudocode | No | The paper describes the proposed method but does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | PRW. PRW contains 11816 surveillance video frames captured by 6 cameras deployed at Tsinghua University, including 43110 pedestrians with 932 different identities. The training set includes 5704 frames, of which there are 15575 pedestrians with 482 identities. The test set contains 450 different identities in 6112 frames and 2057 query probes in total. CUHK-SYSU is another mainstream person search dataset consisting of 18184 street scene photos and movie stills. The training set contains 11206 images and 5532 kinds of pedestrian identities, while the test set uses 2900 queries to search in 6978 gallery images. |
| Dataset Splits | No | The paper specifies training and test sets but does not explicitly mention a separate validation split or how it was used. |
| Hardware Specification | Yes | We implement our method using Py Torch, and all experiments are performed on an NVIDIA Ge Force RTX 3090. |
| Software Dependencies | No | The paper mentions 'Py Torch' but does not specify a version number or other software dependencies with their versions. |
| Experiment Setup | Yes | During training, we employ 4 images resized to 1500 900 as a mini-batch. Our stem network and detection network are optimized by SGD, whose momentum is 0.9, and the initial learning rate is 0.0024. The identification network is Optimized by Adam W with an initial learning rate of 0.0028. For PRW and CUHK-SYSU, the length of the circular queue is set to 5000 and 500, respectively. Besides, random horizontal flip is employed for data augmentation, and tricks such as CWS and PKSample are also adopted. |