Recurrently Aggregating Deep Features for Salient Object Detection
Authors: Xiaowei Hu, Lei Zhu, Jing Qin, Chi-Wing Fu, Pheng-Ann Heng
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We perform experiments to evaluate the effectiveness of the proposed network on 5 famous saliency detection benchmarks and compare it with 15 state-of-the-art methods. Our method ranks first in 4 of the 5 datasets and second in the left dataset. To verify the effectiveness of the proposed RADF model, we evaluate our network equipped with RADF on five famous salient object detection benchmarks, and compare our results against 15 state-of-the-art methods. The experiment results demonstrate that our model quantitatively and qualitatively outperforms others with respect to the accuracy of salient object detection |
| Researcher Affiliation | Academia | 1Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, China 2 Centre for Smart Health, School of Nursing, The Hong Kong Polytechnic University, Hong Kong, China 3 Shenzhen Key Laboratory of Virtual Reality and Human Interaction Technology, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China |
| Pseudocode | No | The paper provides architectural diagrams and mathematical equations but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | https://github.com/xw-hu/RADF |
| Open Datasets | Yes | Our training dataset (MSRA10K dataset (Cheng et al. 2015)) has 10, 000 images with high-quality pixel-wise annotations... ECSSD (Yan et al. 2013) This dataset consists of 1, 000 natural images... HKU-IS (Li and Yu 2015) It includes 4, 447 images... PASCAL-S (Li et al. 2014) This dataset has 850 challenging images... SOD (Martin et al. 2001; Movahedi and Elder 2010) It is composed of 300 images... DUT-OMRON (Yang et al. 2013) This dataset contains 5, 168 high-quality images. |
| Dataset Splits | No | Our network equipped with RADF is trained on the MSRA10K dataset (Cheng et al. 2015) which is widely used for training the salient object detection models (Lee, Tai, and Kim 2016; Zhang et al. 2017a). In addition, images of this dataset are randomly rotated, resized and horizontally flipped for data argumentation, and our model is trained on 4 GPUs with a mini-batch size of 4. |
| Hardware Specification | No | Our model is trained on 4 GPUs with a mini-batch size of 4. |
| Software Dependencies | No | The paper mentions using VGG network, Stochastic Gradient Descent (SGD), and ReLU activation function, but does not provide specific version numbers for any software, libraries, or frameworks used in the implementation. |
| Experiment Setup | Yes | Stochastic gradient descent (SGD) is used to optimize the whole network with the momentum of 0.9 and the weight decay of 0.0005. We set the learning rate as 1e-8 and it reduces by a factor of 0.1 at 7k iterations. Learning stops after 10k iterations. ...our model is trained on 4 GPUs with a mini-batch size of 4. |