Deep Graph Online Hashing for Multi-Label Image Retrieval
Authors: Yuan Cao, Xiangru Chen, Zifan Liu, Wenzhe Jia, Fanlei Meng, Jie Gui
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on two common benchmarks demonstrate that the proposed method achieves up to 13.3% accuracy gains over state-of-the-art baselines and shows competitive performance on training time. |
| Researcher Affiliation | Academia | 1 School of Computer Science and Technology, Ocean University of China, Qingdao, China 2 School of Cyber Science and Engineering, Southeast University, Nanjing, China 3 Engineering Research Center of Blockchain Application, Supervision And Management (Southeast University), Ministry of Education, Nanjing, China 4 Purple Mountain Laboratories, Nanjing, China EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper describes the proposed method, general framework, notations, problem definition, bilayer architecture, loss functions, optimization, and out-of-sample extension using textual descriptions and mathematical formulas, but does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code https://github.com/caoyuan57/DGOH |
| Open Datasets | Yes | MIRFlickr (Huiskes and Lew 2008) consists of 25,000 image instances annotated with 24 semantic labels. MSCOCO (Lin et al. 2014) contains 82,783 training images and 40,504 verification images which belong to 80 categories. |
| Dataset Splits | Yes | MIRFlickr...A random subset of 2,015 instances are selected as the test and the remaining 18,000 instances are used as both the training and the database. MSCOCO...randomly select 2,000 images as the test and the remaining 120,218 images are used as the database. A random subset of 40,000 examples from the database is used for training. |
| Hardware Specification | Yes | The proposed DGOH is trained with Pytorch on a workstation which consists of a CPU with 12 cores, 20 processors and an NVIDIA Geforce RTX 3090. |
| Software Dependencies | No | The proposed DGOH is trained with Pytorch on a workstation which consists of a CPU with 12 cores, 20 processors and an NVIDIA Geforce RTX 3090. (No version for Pytorch is provided.) |
| Experiment Setup | Yes | As for online hash learning, all the images are resized to 224 224 3 and the batch size is set to 50 in the two datasets... We provide the exact values of the parameter configurations in our DGOH as α = β = ηl = 0.01, κ = 1, δ = 0.1, µ = ηf = 0.001, λ = 0.2, n = 200, mt = 100, where ηf and ηl denote the learning rates in the feature network and label network, respectively. |