Improving robustness for vision transformer with a simple dynamic scanning augmentation

Article Computer Science, Artificial Intelligence

CAVER: Cross-Modal View-Mixed Transformer for Bi-Modal Salient Object Detection

Youwei Pang et al.

Summary: Most existing bi-modal salient object detection methods use convolution operation and complex fusion structures. This work proposes a cross-modal view-mixed transformer (CAVER) that aligns and transforms global information. CAVER uses a sequence-to-sequence context propagation and update process with a novel view-mixed attention mechanism. It also simplifies operations with a parameter-free patch-wise token re-embedding strategy. Experimental results show that CAVER surpasses recent state-of-the-art methods on RGB-D and RGB-T SOD datasets when equipped with the proposed components.

IEEE TRANSACTIONS ON IMAGE PROCESSING (2023)

添加到收藏夹

Article Engineering, Electrical & Electronic

No-reference Qquality index of tone-mapped images based on authenticity, preservation, and scene expressiveness

Yang Zhao et al.

Summary: This paper introduces a method for accurately predicting the quality of TMIs: RETI. Based on the characteristics of HDR images, three important elements including authenticity, energy and information preservation, and scene expressiveness are considered, combined with subjective quality for training. The results show that the method has good prediction and generalization abilities compared to some state-of-the-art methods.

SIGNAL PROCESSING (2023)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Efficient data-driven behavior identification based on vision transformers for human activity understanding

Jiachen Yang et al.

Summary: With the development of computer vision, the research on human activity understanding has been greatly promoted. This paper proposes a core weight entropy data information evaluation method based on feature distribution analysis, which effectively reduces data consumption and achieves high performance using a small amount of high information human activity data.

NEUROCOMPUTING (2023)

添加到收藏夹

Article Multidisciplinary Sciences

Adversarial robustness assessment: Why in evaluation both L0 and L∞ attacks are necessary

Shashank Kotyan et al.

Summary: The robustness assessment of machine learning algorithms is a challenging task due to different types of adversarial attacks and defences, as well as the inherent bias in these attacks and defences. This study proposes a model-agnostic adversarial robustness assessment method based on L-0 and L-infinity distance-based norms and robustness levels to address the problems faced. The assessment results show that the robustness may vary significantly depending on the metric used and that L-1 and L-2 metrics alone are not sufficient to avoid spurious adversarial samples. The study also introduces a novel L-infinity black-box adversarial method with lower perturbation than the One-Pixel Attack.

PLOS ONE (2022)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

On the Robustness of Vision Transformers to Adversarial Examples

Kaleel Mahmood et al.

Summary: This study investigates the robustness of Vision Transformers to adversarial examples, finding that these examples do not readily transfer between CNNs and Transformers. The researchers introduce a new attack called the self-attention blended gradient attack and analyze the security of a simple ensemble defense of CNNs and Transformers.

2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021) (2021)

添加到收藏夹

Article Computer Science, Artificial Intelligence