Skip navigation
Please use this identifier to cite or link to this item: https://libeldoc.bsuir.by/handle/123456789/54375
Title: Based on Weak Light YOLOv3 Multi-Target Detection
Authors: Ding Aodi
Lukashevich, P.
Keywords: материалы конференций;yolov3;filter;target detection;residual network
Issue Date: 2023
Publisher: BSU
Citation: Ding Aodi. Based on Weak Light YOLOv3 Multi-Target Detection / Ding Aodi, P. Lukashevich // Pattern Recognition and Information Processing (PRIP'2023) = Распознавание образов и обработка информации (2023) : Proceedings of the 16th International Conference, October 17–19, 2023, Minsk, Belarus / United Institute of Informatics Problems of the National Academy of Sciences of Belarus. – Minsk, 2023. – P. 126–129.
Abstract: Inside the real life scenarios, the YOLOv3 target detection model has achieved good results on many benchmark datasets.the light illumination conditions are poor in many scenarios, such as night, indoor, foggy weather, in dark conditions is still a huge challenge, so in This environment first use the filter to process the image, due to the filter to process high-resolution images is very costly computer resources, so I will use the filter alone to process the filter parameters obtained from high-resolution images transplanted to the original resolution of the image of the model for this experiment, in this experiment choose the detector YOLOv3 as the detection network, YOLOv3 based on the idea of residual network optimization Network multilayer structure can further improve the detection accuracy, especially for small targets, in this yolov3 strengthened the discovery of potentially beneficial information in the image, so the image can be detected in low light with the support of this model framework.
URI: https://libeldoc.bsuir.by/handle/123456789/54375
Appears in Collections:Pattern Recognition and Information Processing (PRIP'2023) = Распознавание образов и обработка информации (2023)

Files in This Item:
File Description SizeFormat 
Ding_Aodi_Based.pdf366.88 kBAdobe PDFView/Open
Show full item record Google Scholar

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.