Huailong Yi, Zhuangzhuang Mao, Mengchao Liu
3D object detection plays an important role in a large number of real-world applications. 3D object detection has achieved high accuracy and efficiency, but the small object detection is still a challenge. In view of the current low detection accuracy of the small objects, the paper studies the multimodal fusion AVOD model to detect cars, pedestrians and cyclists, and fine-tunes the model. The small object detection method based on the skip feature pyramid model is introduced to fuse the detailed information of the multi-layer high-level semantic feature information and the low-level feature map, the object detection accuracy of the model is further improved. The experimental results on the KITTI datasets show that the proposed approach obtains significant improvements.
3D Object Detection; AVOD; feature pyramid; multi-modal fusion; deep convolutional neural network