Human Attention-Guided Explainable AI for Object Detection

Guoyang Liu; Jindi Zhang; Antoni B. Chan; Janet H Hsiao

Human Attention-Guided Explainable AI for Object Detection

Guoyang Liu
, Jindi Zhang
, Antoni B. Chan
, Janet H Hsiao

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

Abstract

Although object detection AI plays an important role in many critical systems, corresponding Explainable AI (XAI) methods remain very limited. Here we first developed FullGrad-CAM and FullGrad-CAM++ by extending traditional gradient-basedmethods to generate object-specific explanations with higher plausibility. Since human attention may reflect features more interpretable to humans, we explored the possibility to use it as guidance to learn how to combine the explanatory information in the detector model to best present as an XAI saliency map that is interpretable (plausible) to humans. Interestingly, we found that human attention maps had higher faithfulness forex plaining the detector model than existing saliency-based XAI methods. By using trainable activation functions and smoothing kernels to maximize the XAI saliency map similarity to human attention maps, the generated map had higher faithfulness and plausibility than both existing XAI methods and human attention maps. The learned functions were model-specific, well generalizable to other databases.

© 2023 The Author(s).

Original language	English
Title of host publication	Proceedings of the 45th Annual Conference of the Cognitive Science Society
Volume	45
Publication status	Published - Jul 2023
Event	The 45th Annual Meeting of the Cognitive Science Society - Darling Harbour, Sydney, Australia Duration: 26 Jul 2023 → 29 Jul 2023 https://cognitivesciencesociety.org/cogsci-2023/

Conference

Conference	The 45th Annual Meeting of the Cognitive Science Society
Place	Australia
City	Sydney
Period	26/07/23 → 29/07/23
Internet address	https://cognitivesciencesociety.org/cogsci-2023/

Bibliographical note

Research Unit(s) information for this publication is provided by the author(s) concerned.

Research Keywords

Object detection
XAI
Human attention
Deep learning
Saliency map

Access Output

https://escholarship.org/uc/item/9r53b44n

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@inproceedings{9513e8d9345c434f8575652553c3754a,

title = "Human Attention-Guided Explainable AI for Object Detection",

abstract = "Although object detection AI plays an important role in many critical systems, corresponding Explainable AI (XAI) methods remain very limited. Here we first developed FullGrad-CAM and FullGrad-CAM++ by extending traditional gradient-basedmethods to generate object-specific explanations with higher plausibility. Since human attention may reflect features more interpretable to humans, we explored the possibility to use it as guidance to learn how to combine the explanatory information in the detector model to best present as an XAI saliency map that is interpretable (plausible) to humans. Interestingly, we found that human attention maps had higher faithfulness forex plaining the detector model than existing saliency-based XAI methods. By using trainable activation functions and smoothing kernels to maximize the XAI saliency map similarity to human attention maps, the generated map had higher faithfulness and plausibility than both existing XAI methods and human attention maps. The learned functions were model-specific, well generalizable to other databases. {\textcopyright} 2023 The Author(s). ",

keywords = "Object detection, XAI, Human attention, Deep learning, Saliency map",

author = "Guoyang Liu and Jindi Zhang and Chan, \{Antoni B.\} and Hsiao, \{Janet H\}",

note = "Research Unit(s) information for this publication is provided by the author(s) concerned.; The 45th Annual Meeting of the Cognitive Science Society ; Conference date: 26-07-2023 Through 29-07-2023",

year = "2023",

month = jul,

language = "English",

volume = "45",

booktitle = "Proceedings of the 45th Annual Conference of the Cognitive Science Society",

url = "https://cognitivesciencesociety.org/cogsci-2023/",

}

TY - GEN

T1 - Human Attention-Guided Explainable AI for Object Detection

AU - Liu, Guoyang

AU - Zhang, Jindi

AU - Chan, Antoni B.

AU - Hsiao, Janet H

N1 - Research Unit(s) information for this publication is provided by the author(s) concerned.

PY - 2023/7

Y1 - 2023/7

N2 - Although object detection AI plays an important role in many critical systems, corresponding Explainable AI (XAI) methods remain very limited. Here we first developed FullGrad-CAM and FullGrad-CAM++ by extending traditional gradient-basedmethods to generate object-specific explanations with higher plausibility. Since human attention may reflect features more interpretable to humans, we explored the possibility to use it as guidance to learn how to combine the explanatory information in the detector model to best present as an XAI saliency map that is interpretable (plausible) to humans. Interestingly, we found that human attention maps had higher faithfulness forex plaining the detector model than existing saliency-based XAI methods. By using trainable activation functions and smoothing kernels to maximize the XAI saliency map similarity to human attention maps, the generated map had higher faithfulness and plausibility than both existing XAI methods and human attention maps. The learned functions were model-specific, well generalizable to other databases. © 2023 The Author(s).

AB - Although object detection AI plays an important role in many critical systems, corresponding Explainable AI (XAI) methods remain very limited. Here we first developed FullGrad-CAM and FullGrad-CAM++ by extending traditional gradient-basedmethods to generate object-specific explanations with higher plausibility. Since human attention may reflect features more interpretable to humans, we explored the possibility to use it as guidance to learn how to combine the explanatory information in the detector model to best present as an XAI saliency map that is interpretable (plausible) to humans. Interestingly, we found that human attention maps had higher faithfulness forex plaining the detector model than existing saliency-based XAI methods. By using trainable activation functions and smoothing kernels to maximize the XAI saliency map similarity to human attention maps, the generated map had higher faithfulness and plausibility than both existing XAI methods and human attention maps. The learned functions were model-specific, well generalizable to other databases. © 2023 The Author(s).

KW - Object detection

KW - XAI

KW - Human attention

KW - Deep learning

KW - Saliency map

M3 - RGC 32 - Refereed conference paper (with host publication)

VL - 45

BT - Proceedings of the 45th Annual Conference of the Cognitive Science Society

T2 - The 45th Annual Meeting of the Cognitive Science Society

Y2 - 26 July 2023 through 29 July 2023

ER -

Human Attention-Guided Explainable AI for Object Detection

Abstract

Conference

Bibliographical note

Research Keywords

Access Output

Other Access Options

Permanent Link

Fingerprint

Cite this