Supervised Matrix Factorization for Cross-Modality Hashing

Authors: Hong Liu, Rongrong Ji, Yongjian Wu, Gang Hua

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the superior performance of SMFH on three cross-modality visual search benchmarks, i.e., the PASCAL-Sentence, Wiki, and NUS-WIDE, with quantitative comparison to various state-of-the-art methods [Kumar and Udupa, 2011; Rastegari et al., 2013; Zhang and Li, 2014; Ding et al., 2014]. We conduct extensive experiments in cross-modality visual search, i.e., using text queries to retrieve relevant images and vice versa, on three widely used benchmarks including, PASCAL-Sentence, Wiki and NUS-WIDE.
Researcher Affiliation Collaboration Fujian Key Laboratory of Sensing and Computing for Smart City, Xiamen University, 361005, China School of Information Science and Engineering, Xiamen University, 361005, China Best Image, Tencent Technology (Shanghai) Co.,Ltd, China [Microsoft Research Asia, Beijing, China
Pseudocode Yes Algorithm 1 Supervised Matrix Factorization Hashing
Open Source Code No The paper states: 'Except CVH, the source codes of the rest methods are available publicly, and all their parameters setting are used as what their papers presented.' This refers to the source code of other methods, not the code for the method proposed in this paper (SMFH).
Open Datasets Yes The PASCAL-Sentence dataset contains 1,000 images... The Wiki dataset contains 2,866 documents... The NUS-WIDE dataset contains 269,648 images... 2http://vision.cs.uiuc.edu/pascal-sentences/ 3http : //www.svcl.ucsd.edu/projects/crossmodal/ 4http : //lms.comp.nus.edu.sg/research/NUSWIDE.htm
Dataset Splits No The paper specifies training and testing splits for each dataset but does not mention a separate validation split.
Hardware Specification Yes All our experiments were run on a workstation with a 3.60GHz Intel Core I5-4790 CPU and 16GB RAM.
Software Dependencies No The paper mentions that 'Except CVH, the source codes of the rest methods are available publicly', referring to other methods' software. It does not provide any specific software dependencies or version numbers for its own implementation.
Experiment Setup Yes Parameter Settings: SMFH has five essential parameters in Eq. (13), i.e., λ1, λ2, , µ, and m. ... we empirically set λ1 = 0.5 for the image modality and λ2 = 0.5 for the text modality. The parameter holds the semantic similarity of the original space, which is set as a large number of 2. µ is a trade-off parameter, which is set as 25 on the two datasets. During each iteration, the number of sampling points is set as 800 for PASCAL-Sentence, 1, 000 for Wiki, and 2, 000 for the NUS-WIDE. ... The three regularization parameters γ, β, and are set to a small number 0.001 in all the experiments.