Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Learning Source-Free Domain Adaptation for Visible-Infrared Person Re-Identification
Authors: Yongxiang Li, Yanglin Feng, Yuan Sun, Dezhong Peng, Xi Peng, Peng Hu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on benchmark datasets demonstrate that SVIP substantially enhances target domain performance and outperforms existing unsupervised VI-Re ID methods under source-free settings. Code is available at https://github.com/LYXRhythm/SVIP. In this paper, we investigate source-free domain adaptation (SFDA) for visibleinfrared person re-identification (VI-Re ID), aiming to adapt a pre-trained source model to an unlabeled target domain without access to source data. |
| Researcher Affiliation | Academia | Yongxiang Li1, Yanglin Feng1, Yuan Sun2, Dezhong Peng1,3, Xi Peng1, Peng Hu1 1College of Computer Science, Sichuan University, Chengdu, China. 2National Key Laboratory of Fundamental Algorithms and Models for Engineering Numerical Simulation, Sichuan University, Chengdu, China. 3Tianfu Jincheng Laboratory, Chengdu, China. EMAIL |
| Pseudocode | No | The paper describes the methodology using mathematical equations and textual descriptions, but does not include any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code is available at https://github.com/LYXRhythm/SVIP. |
| Open Datasets | Yes | Datasets: To systematically investigate domain adaptation in VI-Re ID, we conduct experiments on three widely used datasets: SYSU-MM01 [14], Reg DB [43], and LLCM [44]. |
| Dataset Splits | Yes | The experiment follows the standard evaluations in VI-Re ID [45, 46], including Cumulative Matching Characteristic (CMC), and mean Average Precision (m AP). Domain Adaptation Settings: To comprehensively evaluate the adaptation performance across diverse domains, we define two domain adaptation settings: 1) Basic Setting: This setting is designed to evaluate the model s ability to handle common domain shifts... 2) Weather Setting: To further evaluate model robustness under realistic environmental variations, we introduce five weather conditions to both Reg DB and LLCM... |
| Hardware Specification | Yes | All experiments are performed on an Ubuntu 20.04 system equipped with four NVIDIA RTX 3090 GPUs. |
| Software Dependencies | No | The paper mentions 'Ubuntu 20.04' as the operating system, but does not provide specific version numbers for other key software components, programming languages, or libraries used in the implementation. |
| Experiment Setup | Yes | Implementation Details: During training, pedestrian images are resized to 288 × 144 pixels. The AGW [47] network is employed as the feature extractor. The parameters of the source model remain fixed throughout training. The target model is optimized using the Adam optimizer with a weight decay of 8 × 10−4. The initial learning rate is set to 5 × 10−4 and is reduced by a factor of 10 every 20 epochs. The batch size Nb is 128, and training proceeds for 50 epochs. The momentum parameter η is fixed at 0.1, while the temperature hyperparameter τ is set to 0.05. The clustering algorithm DBSCAN is applied with an epsilon value of 0.6 and a minimum sample size of 4. The threshold T is maintained at 0.5. The trade-off parameters λ1 and λ2 are further analyzed in the Supplementary Material. |