Road Traffic Small Target Detection Algorithm Based on Lightweight YOLO V5
DOI:
https://doi.org/10.54097/fcis.v3i1.6350Keywords:
Deep learning, Lightweight, Small target detectionAbstract
For the current problem of large number of parameters and large computation of the target detection network, the lightness improvement is carried out on the basis of yolov5s, and the dynamic convolution of OD is introduced to compensate for the accuracy degradation caused by the lightness, and the Map0.5 of the improved Yolov5s- EfficientNetV2-OD network reaches 82.8 through comparison experiments, and the FPS is 59 frames per second. The experiment proves that the network maintains the detection accuracy and detection speed of the network for small targets on the basis of light weighting. It is suitable for deployment on vehicle-mounted devices.
Downloads
References
GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation [J]. 2013.
GIRSHICK R. 2015 IEEE International Conference on Computer Vision (ICCV); proceedings of the IEEE International Conference on Computer Vision, F, 2015 [C].
REN S, HE K, GIRSHICK R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks; proceedings of the NIPS, F, 2016 [C].
WEI L, DRAGOMIR A, DUMITRU E, et al. SSD: Single Shot MultiBox Detector [J]. 2016.
REDMON J, DIVVALA S, GIRSHICK R, et al. You Only Look Once: Unified, Real-Time Object Detection; proceedings of the Computer Vision & Pattern Recognition, F, 2016 [C].
REDMON J, FARHADI A. YOLO9000: Better, Faster, Stronger; proceedings of the IEEE Conference on Computer Vision & Pattern Recognition, F, 2017 [C].
REDMON J, FARHADI A J A E-P. YOLOv3: An Incremental Improvement [J]. 2018.
BOCHKOVSKIY A, WANG C Y, LIAO H. YOLOv4: Optimal Speed and Accuracy of Object Detection [J]. 2020.
GE Z, LIU S, WANG F, et al. YOLOX: Exceeding YOLO Series in 2021 [J]. 2021.
REN J C X H, LIU J B, ET AL. Accurate single stage detector using recurrent rolling convolution [J]. IEEE Conference on Computer Vision and Pattern Recognition, 2017, Honolulu, HI, USA. New York: IEEE Press,: 752-60.
AL L X W W W L E. Generalized focal loss： learning qualified and distributed bounding boxes for dense object detection. Advances in Neural Information Processing Systems, 2020.
HUANG WENHAN Y G, GENG KEKE, ET AL. Target detection in complex driving scenes based on adaptive fusion of dilated convolutional features [J]. Journal of Southeast University Natural Science Edition, 2021, (51(06)): 1076-83.
LI CHAO Z A Y A. Omni-dimensional dynamic convolution [EB/OL] https://arxivorg/pdf/220907947pdf, [2022-09-11].
LIN T Y, DOLLáR P, GIRSHICK R, et al. Feature Pyramid Networks for Object Detection [J]. 2016.
LIU S, QI L, QIN H, et al. Path Aggregation Network for Instance Segmentation [J]. 2018.
Q T M L. Efficientnetv2: Smaller models and faster training [J]. International Conference on Machine Learning, 2021: 10096-106.
C T M L Q E R M S F C N N. EfficientNet: Rethinking model scaling for convolutional neural networks [J]. Proceedings of the 36th International Conference on Machine Learning, 2019: 6105-14.r
Downloads
Published
Issue
Section
How to Cite
