Can Eyes on a Car Reduce Traffic Accidents?

Chang, Chia–Ming; Toda, Koki; Gui, Xinyue; Seo, Stela H.; Igarashi, Takeo

doi:10.1145/3543174.3546841

Cited by 22 publications

(1 citation statement)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed method can visualise the areas that the neural network focuses on during inference in a single inference, so it is expected to be suitable for applications such as automated driving and robot control, where realtime performance is required. Recent research has shown that safety can be improved by attaching eyes to vehicles so that automated vehicles and pedestrians can communicate through eye contact [19]. If the proposed method can help the automated driving algorithm to adaptively direct its eyes to the areas where attention is being paid, further safety improvements could be possible.…”

Section: Discussionmentioning

confidence: 99%

Pay Attention via Quantization: Enhancing Explainability of Neural Networks via Quantized Activation

Tashiro

Awano

2023

IEEE Access

View full text Add to dashboard Cite

Modern deep learning algorithms comprise highly complex artificial neural networks, making it extremely difficult for humans to track their inference processes. As the social implementation of deep learning progresses, the human and economic losses caused by inference errors are becoming increasingly problematic, making it necessary to develop methods to explain the basis for the decisions of deep learning algorithms. Although an attention mechanism-based method to visualize the regions that contribute to steering angle prediction in an automated driving task has been proposed, its explanatory capability is low. In this paper, we focus on the fact that the importance of each bit in the activation value of a network is biased (i.e., the sign and exponent bits are weighted more heavily than the mantissa bits), which has been overlooked in previous studies. Specifically, this paper quantizes network activations, encouraging important information to be aggregated to the sign bit. Further, we introduce an attention mechanism restricted to the sign bit to improve the explanatory power. Our numerical experiment using the Udacity dataset revealed that the proposed method achieves a 1.14× higher area under curve (AUC) in terms of the deletion metric.

show abstract

Section: Discussionmentioning

confidence: 99%

Pay Attention via Quantization: Enhancing Explainability of Neural Networks via Quantized Activation

Tashiro

Awano

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

Direct or Indirect: A Video Experiment for In-vehicle Alert Systems

Wang

Chang

Igarashi

2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Gaze-Cues of Humans and Robots on Pedestrian Ways

Jakobowsky,

Abrams,

Rosenthal-von der Pütten

2023

Int J of Soc Robotics

View full text Add to dashboard Cite

Delivery robots and personal cargo robots are increasingly sharing space with incidentally co-present persons (InCoPs) on pedestrian ways facing the challenge of socially adequate and safe navigation. Humans are able to effortlessly negotiate this shared space by signalling their skirting intentions via non-verbal gaze cues. In two online-experiments we investigated whether this phenomenon of gaze cuing can be transferred to human–robot interaction. In the first study, participants (n = 92) watched short videos in which either a human, a humanoid robot or a non-humanoid delivery robot moved towards the camera. In each video, the counterpart looked either straight towards the camera or did an eye movement to the right or left. The results showed that when the counterpart gaze cued to their left, also participants skirted more often to the left from their perspective, thereby walking past each other and avoiding collision. Since the participants were recruited in a right-hand driving country we replicated the study in left-hand driving countries (n = 176). Results showed that participants skirted more often to the right when the counterpart gaze cued to the right, and to the left in case of eye movements to the left, expanding our previous result. In both studies, skirting behavior did not differ regarding the type of counterpart. Hence, gaze cues increase the chance to trigger complementary skirting behavior in InCoPs independently of the robot morphology. Equipping robots with eyes can help to indicate moving direction by gaze cues and thereby improve interactions between humans and robots on pedestrian ways.

show abstract

Can Eyes on a Car Reduce Traffic Accidents?

Cited by 22 publications

References 25 publications

Pay Attention via Quantization: Enhancing Explainability of Neural Networks via Quantized Activation

Pay Attention via Quantization: Enhancing Explainability of Neural Networks via Quantized Activation

Direct or Indirect: A Video Experiment for In-vehicle Alert Systems

Gaze-Cues of Humans and Robots on Pedestrian Ways

Contact Info

Product

Resources

About