機械知覚&ロボティクスグループ
中部大学

Deep Learning Object Detection Point Cloud 口頭発表

LiDAR・カメラのセンサフュージョンによる物体認識モデルの判断根拠の可視化

Author
西尾友佑, 平川翼, 山下隆義, 藤吉弘亘
Publication
自動車技術会秋季大会, 2024

Download: PDF (Japanese)

To realize safe automatic driving, an accurate object detector that can also provide the basis for detection decisions is necessary. Therefore, we adopt a multimodal method, which is currently becoming the mainstream. First, we investigate the influence of each modality on detection results. Furthermore, a perturbation-based basis visualization method for object detection is modified for the multimodal method to show which parts of the image and point cloud are important.

次の研究