Salient Object Detection Based on
Visual Perceptual
Saturation and Two-Stream Hybrid
Networks
Batch No : 16
Guide : Dr.B.Hemantha Kumar Presented By :
M.Rohitha – Y22IT066
K.Ramya Sri – Y22IT055
B.K.N.Sai Lakshmi – Y22IT012
agenda
Abstract
Problem Statement
Key Words
Conclusion
Abstract :
Inspired by the perceived saturation of human visual system, this paper proposes a two-
stream hybrid networks to simulate binocular vision for salient object detection (SOD).
Each stream in our system consists of unsupervised and supervised methods to form a
two-branch module, so as to model the interaction between human intuition and
memory. The two-branch module parallel processes visual information with bottom-up
and top-down SODs, and output two initial saliency maps. Then a polyharmonic neural
network with random-weight (PNNRW) is utilized to fuse two-branch’s perception and
refine the salient objects by learning online via multi-source cues. Depend on visual
perceptual saturation, we can select optimal parameter of superpixel for unsupervised
branch, locate sampling regions for PNNRW, and construct a positive feedback loop to
facilitate perception saturated after the perception fusion. By comparing the binary
outputs of the two-stream, the pixel annotation of predicted object with high
saturation degree could be taken as new training samples. The presented method
constitutes a semi-supervised learning framework actually. Supervised branches only
need to be pre-trained initial, the system can collect the training samples with high
confidence level and then train new models by itself. Extensive experiments show that
the new framework can improve performance of the existing SOD methods, that
exceeds the state-of-the-art methods in six popular benchmarks.
3
Problem statement:
In computer vision and image processing, accurately detecting salient objects is
crucial for numerous applications, including image segmentation, object recognition,
and visual tracking. However, existing methods often struggle to balance precision
and computational efficiency, especially in complex scenes with varied backgrounds
and occlusions.
Traditional approaches may fail to fully leverage the intrinsic visual perceptual cues,
such as color saturation and contrast, which are vital for identifying prominent
objects in an image. Additionally, deep learning-based methods often rely heavily on
a single-stream network architecture, which limits their ability to effectively combine
spatial and contextual information.
This paper addresses the challenge of enhancing the performance of salient object
detection by exploring the role of perceptual saturation and employing a two-stream
hybrid network architecture to integrate low-level visual features with high-level
semantic cues. By tackling these limitations, the proposed method aims to improve
both the accuracy and robustness of salient object detection across diverse visual
scenarios. 4
Key Words
SALIENT OBJECT
TWO-STREAM HYBRID NETWORK
PNNRW
VISUAL PERCEPTUAL SATURATION
Salient Object:
6
A salient object is the most visually prominent element in an
image, standing out due to features like color, contrast, or texture.
It naturally draws attention and is key for tasks like object
recognition and image segmentation. The paper focuses on
detecting these objects using perceptual cues and advanced neural
network methods.
PNNRW :
7
PNNRW - Polyharmonic Neural Network with Random Weights
The PNNRW is used to fuse the outputs of the two branches, combining the
initial saliency maps from both the unsupervised and supervised branches.
This fusion process leverages multi-source cues to refine the salient objects and
enhance the overall accuracy of the detection.
Visual Perceptual Saturation:
8
The concept of visual perceptual saturation plays a crucial role in the
proposed framework. It helps select optimal parameters for the unsupervised
branch, locate sampling regions for the PNNRW, and construct a positive
feedback loop.
This feedback loop facilitates perception saturation after the perception
fusion, ensuring that the system learns from the most salient and informative
regions.
conclusion:
9
The proposed approach to SOD, inspired by the human visual system's
perceived saturation, offers a significant advancement in the field. The two-
stream hybrid network and the use of visual perceptual saturation contribute
to improved accuracy, robustness, and adaptability of the SOD model.
This research paves the way for more sophisticated and efficient SOD
methods, with potential applications in various domains, including image
analysis, video processing, and computer vision.
thank you
PRESENTED BY :
M.ROHITHA – Y22IT066
K.RAMYA SRI – Y22IT055
B.K.N.SAI LAKSHMI – Y22IT012
BATCH NO:16

term pape on salient object detection.pptx

  • 1.
    Salient Object DetectionBased on Visual Perceptual Saturation and Two-Stream Hybrid Networks Batch No : 16 Guide : Dr.B.Hemantha Kumar Presented By : M.Rohitha – Y22IT066 K.Ramya Sri – Y22IT055 B.K.N.Sai Lakshmi – Y22IT012
  • 2.
  • 3.
    Abstract : Inspired bythe perceived saturation of human visual system, this paper proposes a two- stream hybrid networks to simulate binocular vision for salient object detection (SOD). Each stream in our system consists of unsupervised and supervised methods to form a two-branch module, so as to model the interaction between human intuition and memory. The two-branch module parallel processes visual information with bottom-up and top-down SODs, and output two initial saliency maps. Then a polyharmonic neural network with random-weight (PNNRW) is utilized to fuse two-branch’s perception and refine the salient objects by learning online via multi-source cues. Depend on visual perceptual saturation, we can select optimal parameter of superpixel for unsupervised branch, locate sampling regions for PNNRW, and construct a positive feedback loop to facilitate perception saturated after the perception fusion. By comparing the binary outputs of the two-stream, the pixel annotation of predicted object with high saturation degree could be taken as new training samples. The presented method constitutes a semi-supervised learning framework actually. Supervised branches only need to be pre-trained initial, the system can collect the training samples with high confidence level and then train new models by itself. Extensive experiments show that the new framework can improve performance of the existing SOD methods, that exceeds the state-of-the-art methods in six popular benchmarks. 3
  • 4.
    Problem statement: In computervision and image processing, accurately detecting salient objects is crucial for numerous applications, including image segmentation, object recognition, and visual tracking. However, existing methods often struggle to balance precision and computational efficiency, especially in complex scenes with varied backgrounds and occlusions. Traditional approaches may fail to fully leverage the intrinsic visual perceptual cues, such as color saturation and contrast, which are vital for identifying prominent objects in an image. Additionally, deep learning-based methods often rely heavily on a single-stream network architecture, which limits their ability to effectively combine spatial and contextual information. This paper addresses the challenge of enhancing the performance of salient object detection by exploring the role of perceptual saturation and employing a two-stream hybrid network architecture to integrate low-level visual features with high-level semantic cues. By tackling these limitations, the proposed method aims to improve both the accuracy and robustness of salient object detection across diverse visual scenarios. 4
  • 5.
    Key Words SALIENT OBJECT TWO-STREAMHYBRID NETWORK PNNRW VISUAL PERCEPTUAL SATURATION
  • 6.
    Salient Object: 6 A salientobject is the most visually prominent element in an image, standing out due to features like color, contrast, or texture. It naturally draws attention and is key for tasks like object recognition and image segmentation. The paper focuses on detecting these objects using perceptual cues and advanced neural network methods.
  • 7.
    PNNRW : 7 PNNRW -Polyharmonic Neural Network with Random Weights The PNNRW is used to fuse the outputs of the two branches, combining the initial saliency maps from both the unsupervised and supervised branches. This fusion process leverages multi-source cues to refine the salient objects and enhance the overall accuracy of the detection.
  • 8.
    Visual Perceptual Saturation: 8 Theconcept of visual perceptual saturation plays a crucial role in the proposed framework. It helps select optimal parameters for the unsupervised branch, locate sampling regions for the PNNRW, and construct a positive feedback loop. This feedback loop facilitates perception saturation after the perception fusion, ensuring that the system learns from the most salient and informative regions.
  • 9.
    conclusion: 9 The proposed approachto SOD, inspired by the human visual system's perceived saturation, offers a significant advancement in the field. The two- stream hybrid network and the use of visual perceptual saturation contribute to improved accuracy, robustness, and adaptability of the SOD model. This research paves the way for more sophisticated and efficient SOD methods, with potential applications in various domains, including image analysis, video processing, and computer vision.
  • 10.
    thank you PRESENTED BY: M.ROHITHA – Y22IT066 K.RAMYA SRI – Y22IT055 B.K.N.SAI LAKSHMI – Y22IT012 BATCH NO:16