Abstract
Aiming at the problem of small vehicle detection in natural traffic scenes, this paper improves the SSD algorithm. We use K-means algorithm to cluster the Ground Truth of the KITTI dataset to make the default box closer to vehicle. Introduce the FPN framework to achieve the fusion of multi-scale feature maps, and make full use of the feature information contained in deep feature maps and shallow feature maps. And we use Focal Loss to reduce the weight of easy negatives examples, make training focus on hard examples, and improve detection accuracy. Finally, we implement data augmentation for small vehicles so that the training process can extract more small vehicle features to enhance the detection performance. The improved SSD has a mAP of 85.16% on KITTI dataset, which is 6.56% higher than SSD, and an FPS of 43.87, which meets the real-time detection requirements.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
