Lightweight and computationally faster Hypermetropic Convolutional Neural Network for small size object detection

作者：

Highlights：

• A novel CNN architecture to detect small-sized objects is proposed.

• Validation is carried out on various public datasets.

• Results show impressive improvements in detection accuracy and real-time performance.

• It is lighter, smaller and has reduced training time than the state-of-the-art models.

• It is suitable for use in any single-board computer and platforms devoid of GPUs.

摘要

•A novel CNN architecture to detect small-sized objects is proposed.•Validation is carried out on various public datasets.•Results show impressive improvements in detection accuracy and real-time performance.•It is lighter, smaller and has reduced training time than the state-of-the-art models.•It is suitable for use in any single-board computer and platforms devoid of GPUs.

论文关键词：Small-size object detection,Real-time,YOLO,Robotic vision,Faster RCNN,Light-weight models

论文评审过程：Received 27 December 2021, Accepted 25 January 2022, Available online 31 January 2022, Version of Record 7 February 2022.

论文官网地址：https://doi.org/10.1016/j.imavis.2022.104396