Graph-without-cut: An Ideal Graph Learning for Image Segmentation
Authors: Lianli Gao, Jingkuan Song, Feiping Nie, Fuhao Zou, Nicu Sebe, Heng Tao Shen
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive empirical results on three public data sets (i.e, BSDS300, BSDS500 and MSRC) show that our unsupervised GWC achieves state-of-the-art performance compared with supervised and unsupervised image segmentation approaches. We conduct experiments with two validation goals. First, we study the influence of parameters in our algorithm. Second, we compare our results with other state-of-the-art algorithms on three public datasets. |
| Researcher Affiliation | Academia | 1University of Electronic Science and Technology of China, Chengdu 611731, China. 2University of Trento, Trento 38122, Italy. 3Northwestern Polytechnical University, Xi an 710072, China. 4Huazhong University of Science and Technology, China. 5The University of Queensland, Brisbane 4067, Australia. |
| Pseudocode | Yes | Algorithm 1 Solution for ideal graph learning Input: Initialized α, S, segmentation number K, parameters β, γ, μ, a large enough ρ; Output: F, S, α; 1: repeat 2: Fix S and α, calculate F according to the solution of problem (11); 3: Fix F and α, update S by solving the problem (14); 4: Fix F and S, update α by solving the problem (15); 5: until S has K connected components or max iteration is reached. 6: return F, S, α; |
| Open Source Code | No | The paper does not provide any concrete access to source code, such as a specific repository link or an explicit code release statement. |
| Open Datasets | Yes | We experiment with three publicly available data sets for image segmentation: (1) Berkeley Segmentation Data Set 300 (BSDS300) (Martin et al. 2001) consists of all of the grayscale and color segmentations for 300 images. (2) Berkeley Segmentation Data Set 500 (BSDS500) (Arbelaez et al. 2011) is an extension of the BSDS300 with 200 fresh images of the same size. (3) Microsoft Research Cambridge (MSRC) Object Recognition Data Set cleaned version (Malisiewicz and Efros 2007) contains 591 natural images with dense labeling and a large number of object categories and 591 320 213 natural images with single ground truth. |
| Dataset Splits | No | The paper mentions evaluating results at "optimal dataset scale (ODS) and optimal image scale (OIS)" but does not provide specific train/validation/test dataset splits (e.g., percentages or sample counts) needed to reproduce the data partitioning. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details, such as library names with version numbers, needed to replicate the experiment. |
| Experiment Setup | No | The paper discusses the influence of balancing parameters μ and γ, and mentions that μ < 10^2 leads to unsatisfactory performance. However, it does not provide the specific hyperparameter values or training configurations used for generating the main results reported in the tables (Table 1, 2, 3), only a general parameter study. |