Two-Stage Learning to Predict Human Eye Fixations via SDAEs

By Junwei Han, Dingwen Zhang, Shifeng Wen, Lei Guo, Tianming Liu, Xuelong Li. 2016

150 citations recorded by the source corpus.

Saliency detection models aiming to quantitatively predict human eye-attended locations in the visual field have been receiving increasing research interest in recent years. Unlike traditional methods that rely on hand-designed features and contrast inference mechanisms, this paper proposes a novel framework to learn saliency detection models from raw image data using deep networks. The proposed framework mainly consists of two learning stages. At the first learning stage, we develop a stacked denoising autoencoder (SDAE) model to learn robust, representative features from raw image data under an unsupervised manner. The second learning stage aims to jointly learn optimal mechanisms to capture the intrinsic mutual patterns as the feature contrast and to integrate them for final saliency prediction. Given the input of pairs of a center patch and its surrounding patches represented by the features learned at the first stage, a SDAE network is trained under the supervision of eye fixation labels, which achieves both contrast inference and contrast integration simultaneously. Experiments on three publically available eye tracking benchmarks and the comparisons with 16 state-of-the-art approaches demonstrate the effectiveness of the proposed framework.

Subfields: Computer Vision and Pattern Recognition, Sensory Systems, Human-Computer Interaction

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Junwei Han
2. Gong Cheng
3. Liangpei Zhang
4. Peicheng Zhou
5. Bo Du
6. Dacheng Tao
7. Xiaoqiang Lu
8. Shuhui Bu

In Sensory Systems

1. Gong Cheng
2. L. Jiao
3. Min Wang
4. Shuyuan Yang
5. Shigang Wang
6. Junwei Han

In Human-Computer Interaction

1. Dacheng Tao
2. L. Jiao
3. Min Wang
4. Shuyuan Yang
5. Gong Cheng
6. Shigang Wang
7. Junwei Han
8. Xiaohui Liu

Citations in the corpus

• Stacked Convolutional Denoising Auto-Encoders for Feature Representation (2017, background): “given thorough proof in theory and in practice [23], [37].”
• Scene parsing using inference Embedded Deep Networks (2016, background): “[13,14] are explored through simulating human vision mechanism,”
• New Contour Cue-Based Hybrid Sparse Learning for Salient Object Detection (2019, background): “Models in this kind mostly explore the sparsity/singularity, local/global contrast, psychological principles, statistical priors, and learning strategies for gaze prediction [9]–[14].”
• Revealing Event Saliency in Unconstrained Video Collection (2017, unknown): “To avoid overfitting, we follow [44] to divide each of the video collection into non-overlapped 10 parts.”
• Learning Rotation-Invariant Convolutional Neural Networks for Object Detection in VHR Optical Remote Sensing Images (2016, background): “Since then, deep learning algorithms particularly convolutional neural networks (CNNs) have shown their stronger feature representation power in a wide range of computer vision applications [41]–[46].”
• Hierarchical and Robust Convolutional Neural Network for Very High-Resolution Remote Sensing Object Detection (2019, unknown): “tasks [3], [40]–[44].”
• Weakly supervised target detection in remote sensing images based on transferred deep features and negative bootstrapping (2015, methodology): “Finally, a post-processing scheme is used to eliminate repeated detections via non-maximum suppression (Cheng et al. 2013b, 2014; Han et al. 2015a, 2014; Zhang et al. 2015).”
• A survey of deep neural network architectures and their applications (2017, background): “The deep learning techniques only require raw image data [52].”
• An Intelligent Fault Diagnosis Method Using Unsupervised Feature Learning Towards Mechanical Big Data (2016, background): “of softmax regression to take values close to zero while permits other parameters to retain their relatively large values, thereby improving the generalization ability [35].”