Robust Self-Paced Hashing for Cross-Modal Retrieval with Noisy Labels

Authors: Ruitao Pu, Yuan Sun, Yang Qin, Zhenwen Ren, Xiaomin Song, Huiming Zheng, Dezhong Peng

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate that the proposed RSHNL performs remarkably well over the state-of-the-art CMH methods. To verify the effectiveness of the proposed RSHNL, we conduct extensive experiments on four widely used datasets, i.e., XMedia (Peng et al. 2015), INRIA-Websearch (Krapac et al. 2010), Wikipedia (Rasiwasia et al. 2010), and XMedia Net (Peng, Huang, and Zhao 2018).
Researcher Affiliation Collaboration 1Sichuan University, Chengdu, China, 610044, 2Southwest University of Science and Technology, Mianyang, China, 621010, 3Sichuan National Innovation New Vision UHD Video Technology Co., Ltd., Chengdu, China, 610095
Pseudocode No The training process of RSHNL is shown in the appendix. (No pseudocode or algorithm block is present in the main text of the paper.)
Open Source Code Yes Code https://github.com/perquisite/RSHNL
Open Datasets Yes Dataset To verify the effectiveness of the proposed RSHNL, we conduct extensive experiments on four widely used datasets, i.e., XMedia (Peng et al. 2015), INRIA-Websearch (Krapac et al. 2010), Wikipedia (Rasiwasia et al. 2010), and XMedia Net (Peng, Huang, and Zhao 2018).
Dataset Splits No The paper mentions using a "testing set" and "validation set" but does not provide specific percentages, sample counts, or explicit splitting methodology for these datasets in the main text. More details on implementation are provided in the appendix due to space limitations.
Hardware Specification Yes Besides, all experiments are conducted on a single Ge Force RTX3090Ti 24GB GPU and our RSHNL is implemented in Py Torch 1.12.0.
Software Dependencies Yes our RSHNL is implemented in Py Torch 1.12.0.
Experiment Setup No where α is a hyper-parameter, t is the current epoch, Nw and Nm are warm-up epoch and maximal epoch, respectively. (The paper mentions hyperparameters like alpha, Nw, and Nm but does not provide their specific values in the main text. It states that more implementation details are in the appendix.)