Salient Objects in Clutter

Abstract

This paper identifies and addresses a serious design bias of existing salient object detection (SOD) datasets, which unrealistically assume that each image should contain at least one clear and uncluttered salient object. This design bias has led to a saturation in performance for state-of-the-art SOD models when evaluated on existing datasets. However, these models are still far from satisfactory when applied to real-world scenes. Based on our analyses, we propose a new high-quality dataset and update the previous saliency benchmark. Specifically, our dataset, called Salient Objects in Clutter~\textbf{(SOC)}, includes images with both salient and non-salient objects from several common object categories. In addition to object category annotations, each salient image is accompanied by attributes that reflect common challenges in common scenes, which can help provide deeper insight into the SOD problem. Further, with a given saliency encoder, e.g., the backbone network, existing saliency models are designed to achieve mapping from the training image set to the training ground-truth set. We, therefore, argue that improving the dataset can yield higher performance gains than focusing only on the decoder design. With this in mind, we investigate several dataset-enhancement strategies, including label smoothing to implicitly emphasize salient boundaries, random image augmentation to adapt saliency models to various scenarios, and self-supervised learning as a regularization strategy to learn from small datasets. Our extensive results demonstrate the effectiveness of these tricks. We also provide a comprehensive benchmark for SOD, which can be found in our repository: https://github.com/DengPingFan/SODBenchmark.

349 references, 20 pages, survey 201 models, benchmark 100 models. Online benchmark: https://github.com/DengPingFan/SODBenchmark

References in corpus (21)

• Distilling the Knowledge in a Neural Network
• Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials
• U${}^2$-Net: Going Deeper with Nested U-Structure for Salient Object Detection
• On Calibration of Modern Neural Networks
• Concealed Object Detection
• Dense Attention Fluid Network for Salient Object Detection in Optical Remote Sensing Images
• RGB-D Salient Object Detection: A Survey
• Delving Deep into Label Smoothing
• Dynamic Feature Integration for Simultaneous Detection of Salient Object, Edge and Skeleton
• Re-thinking Co-Salient Object Detection
• Light Field Salient Object Detection: A Review and Benchmark
• Boundary-Aware Segmentation Network for Mobile and Web Applications
• Salient Object Detection with Lossless Feature Reflection and Weighted Structural Loss
• Regularized Densely-connected Pyramid Network for Salient Instance Segmentation
• Altitude Training: Strong Bounds for Single-Layer Dropout
• SVAM: Saliency-guided Visual Attention Modeling by Autonomous Underwater Robots
• Richer and Deeper Supervision Network for Salient Object Detection
• Deep Reasoning with Multi-Scale Context for Salient Object Detection
• IDA: Improved Data Augmentation Applied to Salient Object Detection
• BiconNet: An Edge-preserved Connectivity-based Approach for Salient Object Detection
• Salient Image Matting

Cited by in corpus (6)

• Fast Camouflaged Object Detection via Edge-based Reversible Re-calibration Network
• Video Polyp Segmentation: A Deep Learning Perspective
• Advances in Deep Concealed Scene Understanding
• Bilateral Reference for High-Resolution Dichotomous Image Segmentation
• View-aware Salient Object Detection for 360° Omnidirectional Image
• Uncertainty Guided Refinement for Fine-Grained Salient Object Detection