Deep Region Hashing for Generic Instance Search from Images
Authors: Jingkuan Song, Tao He, Lianli Gao, Xing Xu, Heng Tao Shen
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on four datasets show that our DRH can achieve even better performance than the state-of-the-arts in terms of m AP, while the efficiency is improved by nearly 100 times. |
| Researcher Affiliation | Academia | Jingkuan Song, Tao He, Lianli Gao, Xing Xu, Heng Tao Shen Center for Future Media and School of Computer Science and Engineering, University of Electronic Science and Technology of China. {jingkuan.song, hetaoconquer}@gmail.com, {lianli.gao,xing.xu}@uestc.edu.cn, shenhengtao@hotmail.com |
| Pseudocode | No | The paper 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 | Datasets and Evaluation Metric We consider two publicly available datasets. 1) Oxford Buildings (Philbin et al. 2007) contains two sub-datasets: Oxford 5k... 2) Paris Buildings (Philbin et al. 2008) contains two sub-datasets as well: Paris 6k... The hashing layer is trained on a dataset which is composed of Oxford 5K training examples and Paris 6k training examples. |
| Dataset Splits | No | The paper mentions training and testing datasets, but does not provide specific details on how these datasets were split into training, validation, and test sets with exact percentages or sample counts for reproducibility, nor does it cite predefined validation splits. |
| Hardware Specification | Yes | All the experiments are done on a server with Intel@Core i7-6700K. The graphics is Ge Gorce GTX TITAN X/PCle/SSE2. |
| Software Dependencies | No | The paper mentions 'CPP' for implementation but does not specify any other software dependencies, libraries, or frameworks with version numbers that would be needed to replicate the experiment. |
| Experiment Setup | Yes | The shared convolutional layers are initialized by pretraining a model (Radenovic, Tolias, and Chum 2016). Next, we tune all the layers of the Ro I layer and the hashing layer to conserve memory. In particular, for RPN training we fix the convolutional layers and only update the RPN layers parameters. ...learning rate is set to 0.01 in the experiments. The parameters α, β, η were empirically set as 100, 0.001 and 0.001 respectively. ... For RPN, we choose 300 regions to generate region hash code for l DRH. For sliding window, we tune λ = [0.4, 0.5, 0.6, 0.7] to get different number of object proposals. ... When M = 400, the performance is 0.3% lower than M = 800. Considering the memory cost, we choose M = 400 for the following experiments. ... In the following experiments, q is set as 6. |