An Improved Deep Learning Solution for Object Detection in Self-Driving Cars

Engineering notes

In order to train and evaluate the neural network, we used BDD100K dataset which is one of the largest open-source datasets in autonomous driving published by Berkeley University.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Reliable object detection is one of the most important requirements of environment perception in autonomous driving. The goal of this research is to find a convenient solution to detect objects in images from the self-driving car medium. Convolutional neural networks (CNNs) are deep neural networks used in image processing, object classification, and object recognition. Therefore, deep convolution networks are employed in this project to identify objects accurately. In order to train and evaluate the neural network, we used BDD100K dataset which is one of the largest open-source datasets in autonomous driving published by Berkeley University. The approach used in the proposed algorithm is to apply the feature pyramid network along with a single-stage object detector, which enhances the accuracy of object detection. In addition, it improves the detection of different scales, especially small ones compared to those of the previous works, leading to increased safety and security in self-driving cars.

Links and sources

• Official / arXiv page

Comments