Multi-level Net: A Visual Saliency Prediction Model
نویسندگان
چکیده
• We use three popular CNN models: VGG16, VGG19 and AlexNet. • To limit rescaling, the last pooling stage is removed and the stride of the last but one pooling layer is decreased. • We take feature maps at three different locations of the FCN, and concatenate them to form a single tensor. • A 3 x 3 convolutional layer learns 64 saliency-specific feature maps, then a 1 x 1 convolution learns to weight each map to produce a temporary saliency prediction.
منابع مشابه
Compressed-Sampling-Based Image Saliency Detection in the Wavelet Domain
When watching natural scenes, an overwhelming amount of information is delivered to the Human Visual System (HVS). The optic nerve is estimated to receive around 108 bits of information a second. This large amount of information can’t be processed right away through our neural system. Visual attention mechanism enables HVS to spend neural resources efficiently, only on the selected parts of the...
متن کاملGraph-based Visual Saliency Model using Background Color
Visual saliency is a cognitive psychology concept that makes some stimuli of a scene stand out relative to their neighbors and attract our attention. Computing visual saliency is a topic of recent interest. Here, we propose a graph-based method for saliency detection, which contains three stages: pre-processing, initial saliency detection and final saliency detection. The initial saliency map i...
متن کاملJust Noticeable Difference Estimation Using Visual Saliency in Images
Due to some physiological and physical limitations in the brain and the eye, the human visual system (HVS) is unable to perceive some changes in the visual signal whose range is lower than a certain threshold so-called just-noticeable distortion (JND) threshold. Visual attention (VA) provides a mechanism for selection of particular aspects of a visual scene so as to reduce the computational loa...
متن کاملA Saliency Detection Model via Fusing Extracted Low-level and High-level Features from an Image
Saliency regions attract more human’s attention than other regions in an image. Low- level and high-level features are utilized in saliency region detection. Low-level features contain primitive information such as color or texture while high-level features usually consider visual systems. Recently, some salient region detection methods have been proposed based on only low-level features or hig...
متن کاملSaliency-based Sequential Image Attention with Multiset Prediction
Humans process visual scenes selectively and sequentially using attention. Central to models of human visual attention is the saliency map. We propose a hierarchical visual architecture that operates on a saliency map and uses a novel attention mechanism to sequentially focus on salient regions and take additional glimpses within those regions. The architecture is motivated by human visual atte...
متن کامل