Axiom-based Grad-CAM: Towards Accurate Visualization and Explanation of CNNs

Fu, Ruigang; Hu, Qingyong; Dong, Xiaohu; Guo, Yulan; Gao, Yinghui; Li, Biao

doi:10.48550/arxiv.2008.02312

Cited by 62 publications

(94 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In this section, we also compare two more advanced visualization methods, i.e., XGrad-CAM [56] and Grad-CAM++ [55], to verify the effectiveness of Grad-CAM [44] used in our method for affordance grounding both subjectively and objectively. − XGrad-CAM [56]: This paper introduces two axioms: sensitivity [57] and consistency [58], and makes XGrad-CAM satisfy these two constraints. XGrad-CAM is a visualization method with class discriminative ability to highlight the relevant regions belonging to the class, which is calculated as follows:…”

Section: Ablation Studiesmentioning

confidence: 99%

Learning Visual Affordance Grounding from Demonstration Videos

Luo,

Zhai,

Zhang

et al. 2021

Preprint

View full text Add to dashboard Cite

Section: Ablation Studiesmentioning

confidence: 99%

Learning Visual Affordance Grounding from Demonstration Videos

Luo,

Zhai,

Zhang

et al. 2021

Preprint

View full text Add to dashboard Cite

“…Following this hypothesis, we extract the saliency maps for the ImageNet validation set with the XGradCAM [24] approach, using the ResNet-50 [31] model with true category labels obtained from official code repository. Fig.…”

Section: G Visualizing Saliency Maps and Patchrankmentioning

confidence: 99%

PatchGame: Learning to Signal Mid-level Patches in Referential Games

Gupta¹,

Somepalli²,

Gupta³

et al. 2021

Preprint

View full text Add to dashboard Cite

We study a referential game (a type of signaling game) where two agents communicate with each other via a discrete bottleneck to achieve a common goal. In our referential game, the goal of the speaker is to compose a message or a symbolic representation of "important" image patches, while the task for the listener is to match the speaker's message to a different view of the same image. We show that it is indeed possible for the two agents to develop a communication protocol without explicit or implicit supervision. We further investigate the developed protocol and show the applications in speeding up recent Vision Transformers by using only important patches, and as pre-training for downstream recognition tasks (e.g., classification).Code is available at https://kampta.github.io/patch-game.

show abstract

“…To overcome this drawback, Gradient-weighted CAM (Grad-CAM) was proposed which can generate high quality saliency maps on the original network architecture without any modification [17]. Recent extensions of Grad-CAM such as Grad-CAM++ [18], XGrad-CAM [19], Eigen-CAM [20] are also proved to be powerful XAI tools for visualizing the class-specific decision that deep neural networks make.…”

Section: Saliency Maps As Xai Toolsmentioning

confidence: 99%

“…Regions of saliency in an image are those largely relevant to the decision that an AI model made. Various methods have been developed for creating high-quality, easily interpretable saliency maps [12,13,14,15,16,17,18,19,20]. While these studies have built a methodological foundation for XAI in autonomous driving, they are difficult to be evaluated.…”

Section: Introductionmentioning

confidence: 99%

Towards explainable artificial intelligence (XAI) for early anticipation of traffic accidents

Karim

Qin

2021

Preprint

View full text Add to dashboard Cite

Traffic accident anticipation is a vital function of Automated Driving Systems (ADSs) for providing a safety-guaranteed driving experience. An accident anticipation model aims to predict accidents promptly and accurately before they occur. Existing Artificial Intelligence (AI) models of accident anticipation lack a human-interpretable explanation of their decision-making. Although these models perform well, they remain a black-box to the ADS users, thus difficult to get their trust. To this end, this paper presents a Gated Recurrent Unit (GRU) network that learns spatio-temporal relational features for the early anticipation of traffic accidents from dashcam video data. A post-hoc attention mechanism named Grad-CAM is integrated into the network to generate saliency maps as the visual explanation of the accident anticipation decision. An eye tracker captures human eye fixation points for generating human attention maps. The explainability of network-generated saliency maps is evaluated in comparison to human attention maps. Qualitative and quantitative results on a public crash dataset confirm that the proposed explainable network can anticipate an accident on average 4.57 seconds before it occurs, with 94.02% average precision. In further, various post-hoc attention-based XAI methods are evaluated and compared. It confirms that the Grad-CAM chosen by this study can generate high-quality, human-interpretable saliency maps (with 1.42 Normalized Scanpath Saliency) for explaining the crash anticipation decision. Importantly, results confirm that the proposed AI model, with a human-inspired design, can outperform humans in the accident anticipation.

show abstract

Axiom-based Grad-CAM: Towards Accurate Visualization and Explanation of CNNs

Cited by 62 publications

References 8 publications

Learning Visual Affordance Grounding from Demonstration Videos

Learning Visual Affordance Grounding from Demonstration Videos

PatchGame: Learning to Signal Mid-level Patches in Referential Games

Towards explainable artificial intelligence (XAI) for early anticipation of traffic accidents

Contact Info

Product

Resources

About