Online Dictionary Learning on Symmetric Positive Definite Manifolds with Vision Applications
Authors: Shengping Zhang, Shiva Kasiviswanathan, Pong Yuen, Mehrtash Harandi
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical results on both large-scale image classification task and dynamic video processing tasks validate the superior performance of our approach as compared to several state-of-the-art algorithms. |
| Researcher Affiliation | Collaboration | 1School of Computer Science and Technology, Harbin Institute of Technology at Weihai, China 2General Electric Global Research, United States 3Department of Computer Science, Hong Kong Baptist University, Hong Kong 4NICTA and Australian National University, Australia |
| Pseudocode | Yes | Algorithm 1: : Kernel sparse coding (1) using APG and Algorithm 2: : Online Framework for Sparse Coding and Dictionary Learning on SPD Manifold |
| Open Source Code | No | The paper mentions 'Codes of all compared methods were downloaded from respective authors websites' and that 'the code is in Matlab' for their own approach, but does not explicitly state that their own code is open-source or provide a link. |
| Open Datasets | Yes | We used 10 sequences from two popular benchmark datasets: Wallflower3 and I2R4 for comparison, which contain a variety of challenges such as illumination variation or dynamic backgrounds. We chose two handwritten digit datasets MNIST (Lecun et al. 1998) and USPS5 for testing. |
| Dataset Splits | Yes | For each video sequence, we used the first 200 frames not containing the foreground objects for training and the remaining frames were used for detection (testing). The MNIST dataset has 60000 training and 10000 testing images, whereas the USPS dataset has 7291 training and 2007 testing images. |
| Hardware Specification | Yes | All experiments were done on a PC with 2.83GHz CPU and 6GB memory. |
| Software Dependencies | No | The paper mentions that its code is in Matlab but does not provide specific version numbers for Matlab or any other ancillary software libraries or packages. |
| Experiment Setup | Yes | The values of algorithmic parameters used in our experiments are: k = 5, λ = 0.01, T = 5, β = 0.1, and α = 10-7. These parameter were tuned and then fixed for all experiments. |