A REVIEW OF DEEP LEARNING-BASED MULTI-OBJECT TRACKING METHODS

Mirzaaxmedov , Dilmurod

Рақамли иқтисодиёт ва ахборот технологиялари · 2026-yil

Annotatsiya

This paper reviews ten recent influential studies addressing the Multiple Object Tracking (MOT) problem. The analysis focuses on object detection, reidentification, anchor-free methods, deep learning-based approaches, and real-time tracking performance. MOT is a key task in computer vision, aimed at detecting and tracking multiple objects in real-time video streams, with important applications in traffic management, surveillance, healthcare monitoring, and agriculture. Recent advances in deep learning and multi-task learning have significantly improved tracking accuracy and robustness. The reviewed works examine various tracking frameworks, including Siamese-based models and neural network architectures such as CNNs and RNNs. A comparative evaluation is also conducted using a column diagram to assess the effectiveness of different MOT methods, including FairMOT, YOLO, CenterNet, and classical algorithms such as the Kalman filter and Hungarian algorithm

Maqola ma’lumotlari
MualliflarMirzaaxmedov , Dilmurod
JurnalРақамли иқтисодиёт ва ахборот технологиялари
Nashr sanasi2026-04-14
Jild6
Son1
Betlar161-167
TilIngliz

Maqolaning asl sahifasi

Kalit so‘zlar

Multi-Object Tracking, Re-Identification, Deep Learning, Convolutional Neural Networks, Siamese Networks, Real-Time Object Tracking, Object Detection, Trajectory Prediction, Feature Extraction, Online Tracking

Ilmiy soha

Iqtisodiyot

Рақамли иқтисодиёт ва ахборот технологиялари jurnalidan boshqa maqolalar

Рақамли иқтисодиёт ва ахборот технологиялари — barcha maqolalar