Skip to content
DSML.KZ
Publications

Talk · International Conference on Machine Learning and Cybernetics · July 12, 2025

SEPE-YOLO: A Squeeze-and-Excitation Aided and Parzen Estimator Optimized Yolo Model for Vehicle Detection

Sourajit Maity, Sayani Dhar, Asfak Ali, Dmitrii Kaplun, Sergei Antonov, Ram Sarkar

Dmitry Kaplun
Open via DOI

Abstract

Vehicle detection systems (VDS) are growing in popularity as intelligent transportation systems expand rapidly. Autonomous driving, urban planning, and traffic regulation all depend on VDS. This paper develops a dependable VDS model, called SEPE-YOLO (Squeeze-and-Excitation aided and Parzen Estimator optimized YOLO), which can detect various vehicle types from images taken in different lighting and weather conditions, including sunny, rainy, daylight, and midnight. In order to improve the model’s performance, Squeeze-and-Excitation (SE) blocks are incorporated into the YOLOv10 architecture. We have used the Tree-structured Parzen Estimator to fine-tune the hyperparameters of the SEPE-YOLO model. The proposed model has been evaluated on two public VDS datasets: JUVDsiv1 and IRUVD. We have obtained mAP scores of 79.9% and 94.8% on JUVDsiv1 and IRUVD dataets, respectively.

Advanced Neural Network ApplicationsAutonomous Vehicle Technology and SafetyDomain Adaptation and Few-Shot Learningvehicle detectionYOLOv10squeeze-and-excitation blocks

More by Dmitry Kaplun