Eye-Contact Transformer: Eye-Contact Detection of Distant Pedestrians Considering Scene Context
-
- HATA Ryusei
- Nagoya Univ.
-
- DEGUCHI Daisuke
- Nagoya Univ.
-
- HIRAYAMA Takatsugu
- Nagoya Univ.,Univ. of Human Environments
-
- KAWANISHI Yasutomo
- Nagoya Univ.,RIKEN
-
- MURASE Hiroshi
- Nagoya Univ.
Bibliographic Information
- Other Title
-
- Eye-contact Transformer: シーンコンテクストを考慮した遠方歩行者のアイコンタクト検出
Abstract
It is very important for safe driving to understand whether a pedestrian is aware of our own vehicle. Eye-contact plays an important role in determining the awareness of each pedestrian. Traditionally, eye gaze estimation has been widely used for such eye contact detection, but this method cannot be applied in road environments since the distance between pedestrians and vehicles is large. Therefore, this paper proposes a method that can detect eye-contact from pedestrians accurately even in complex traffic environments by capturing the relationship between the sequence of pedestrian's skeleton and scene context of the surrounding environment based on Transformer-based model. Through the experiments using in-vehicle camera image dataset, we confirmed the effectiveness of the proposed method.
Journal
-
- 電子情報通信学会論文誌D 情報・システム
-
電子情報通信学会論文誌D 情報・システム J107-D (4), 231-242, 2024-04-01
The Institute of Electronics, Information and Communication Engineers
- Tweet
Details 詳細情報について
-
- CRID
- 1390299595849886592
-
- ISSN
- 18810225
- 18804535
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
-
- Abstract License Flag
- Disallowed