車載カメラ映像からの物体検出における半教師あり学習の有効性の検証
- Author
- 法華津伸一，山下隆義, 藤吉弘亘, 平川翼
- Publication
- 自動車技術会春季大会, 2023
Download: PDF (Japanese)
In automated driving technology, it is essential to understand environmental information such as the location of surrounding vehicles and
pedestrians. Object detection is a technique for acquiring environmental information. Object detection is the task of detecting the position and
class of objects in an image. Object detection can be made more accurate by learning with a large amount of supervised data. However, the
high cost of annotating the data makes it difficult to create large supervised datasets. For this reason, research using semi-supervised learning
for object detection has been attracting attention in recent years. Previous studies of semi-supervised learning in object detection tasks have
mainly conducted evaluation experiments only on large datasets with many classes, such as MS COCO and PASCAL VOC. Therefore, the
effectiveness of semi-supervised learning in the case of in-vehicle camera data as input has not yet been demonstrated. In this study, we
examine the effectiveness of semi-supervised learning in object detection when in-vehicle camera data is used as input. We also consider the
class imbalance problem and propose to add class weights to the class classification loss. Experimental results using BDD100K show that
semi-supervised learning is also effective when in-vehicle camera images are used as input. We also confirmed that the proposed method
improves the class imbalance problem and accuracy.