TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Disentangled High Quality Salient Object Detection

Disentangled High Quality Salient Object Detection

Lv Tang, Bo Li, Shouhong Ding, Mofei Song

2021-08-08ICCV 2021 10regressionVocal Bursts Intensity PredictionSalient Object Detectionobject-detectionObject DetectionRGB Salient Object Detection
PaperPDFCodeCode(official)

Abstract

Aiming at discovering and locating most distinctive objects from visual scenes, salient object detection (SOD) plays an essential role in various computer vision systems. Coming to the era of high resolution, SOD methods are facing new challenges. The major limitation of previous methods is that they try to identify the salient regions and estimate the accurate objects boundaries simultaneously with a single regression task at low-resolution. This practice ignores the inherent difference between the two difficult problems, resulting in poor detection quality. In this paper, we propose a novel deep learning framework for high-resolution SOD task, which disentangles the task into a low-resolution saliency classification network (LRSCN) and a high-resolution refinement network (HRRN). As a pixel-wise classification task, LRSCN is designed to capture sufficient semantics at low-resolution to identify the definite salient, background and uncertain image regions. HRRN is a regression task, which aims at accurately refining the saliency value of pixels in the uncertain region to preserve a clear object boundary at high-resolution with limited GPU memory. It is worth noting that by introducing uncertainty into the training process, our HRRN can well address the high-resolution refinement task without using any high-resolution training data. Extensive experiments on high-resolution saliency datasets as well as some widely used saliency benchmarks show that the proposed method achieves superior performance compared to the state-of-the-art methods.

Results

TaskDatasetMetricValueModel
Object DetectionDAVIS-SF-measure0.935Tang et al.
Object DetectionDAVIS-SMAE0.012Tang et al.
Object DetectionDAVIS-SS-measure0.92Tang et al.
Object DetectionDAVIS-SmBA0.716Tang et al.
Object DetectionHRSODMAE0.022Tang et al. (HRSOD)
Object DetectionHRSODS-Measure0.92Tang et al. (HRSOD)
Object DetectionHRSODmBA0.693Tang et al. (HRSOD)
Object DetectionHRSODmax F-Measure0.915Tang et al. (HRSOD)
3DDAVIS-SF-measure0.935Tang et al.
3DDAVIS-SMAE0.012Tang et al.
3DDAVIS-SS-measure0.92Tang et al.
3DDAVIS-SmBA0.716Tang et al.
3DHRSODMAE0.022Tang et al. (HRSOD)
3DHRSODS-Measure0.92Tang et al. (HRSOD)
3DHRSODmBA0.693Tang et al. (HRSOD)
3DHRSODmax F-Measure0.915Tang et al. (HRSOD)
RGB Salient Object DetectionDAVIS-SF-measure0.935Tang et al.
RGB Salient Object DetectionDAVIS-SMAE0.012Tang et al.
RGB Salient Object DetectionDAVIS-SS-measure0.92Tang et al.
RGB Salient Object DetectionDAVIS-SmBA0.716Tang et al.
RGB Salient Object DetectionHRSODMAE0.022Tang et al. (HRSOD)
RGB Salient Object DetectionHRSODS-Measure0.92Tang et al. (HRSOD)
RGB Salient Object DetectionHRSODmBA0.693Tang et al. (HRSOD)
RGB Salient Object DetectionHRSODmax F-Measure0.915Tang et al. (HRSOD)
2D ClassificationDAVIS-SF-measure0.935Tang et al.
2D ClassificationDAVIS-SMAE0.012Tang et al.
2D ClassificationDAVIS-SS-measure0.92Tang et al.
2D ClassificationDAVIS-SmBA0.716Tang et al.
2D ClassificationHRSODMAE0.022Tang et al. (HRSOD)
2D ClassificationHRSODS-Measure0.92Tang et al. (HRSOD)
2D ClassificationHRSODmBA0.693Tang et al. (HRSOD)
2D ClassificationHRSODmax F-Measure0.915Tang et al. (HRSOD)
2D Object DetectionDAVIS-SF-measure0.935Tang et al.
2D Object DetectionDAVIS-SMAE0.012Tang et al.
2D Object DetectionDAVIS-SS-measure0.92Tang et al.
2D Object DetectionDAVIS-SmBA0.716Tang et al.
2D Object DetectionHRSODMAE0.022Tang et al. (HRSOD)
2D Object DetectionHRSODS-Measure0.92Tang et al. (HRSOD)
2D Object DetectionHRSODmBA0.693Tang et al. (HRSOD)
2D Object DetectionHRSODmax F-Measure0.915Tang et al. (HRSOD)
16kDAVIS-SF-measure0.935Tang et al.
16kDAVIS-SMAE0.012Tang et al.
16kDAVIS-SS-measure0.92Tang et al.
16kDAVIS-SmBA0.716Tang et al.
16kHRSODMAE0.022Tang et al. (HRSOD)
16kHRSODS-Measure0.92Tang et al. (HRSOD)
16kHRSODmBA0.693Tang et al. (HRSOD)
16kHRSODmax F-Measure0.915Tang et al. (HRSOD)

Related Papers

Language Integration in Fine-Tuning Multimodal Large Language Models for Image-Based Regression2025-07-20A Real-Time System for Egocentric Hand-Object Interaction Detection in Industrial Domains2025-07-17RS-TinyNet: Stage-wise Feature Fusion Network for Detecting Tiny Objects in Remote Sensing Images2025-07-17Decoupled PROB: Decoupled Query Initialization Tasks and Objectness-Class Learning for Open World Object Detection2025-07-17Dual LiDAR-Based Traffic Movement Count Estimation at a Signalized Intersection: Deployment, Data Collection, and Preliminary Analysis2025-07-17Neural Network-Guided Symbolic Regression for Interpretable Descriptor Discovery in Perovskite Catalysts2025-07-16Imbalanced Regression Pipeline Recommendation2025-07-16Second-Order Bounds for [0,1]-Valued Regression via Betting Loss2025-07-16