Empirical Study of the Performance of Object Detection Methods on Road Marking Dataset

Vo, Nguyen D.; Le, Vy; Lai, Thien C.; Duong, Khanh B. T.; Tran, Yen

doi:10.1109/rivf55975.2022.10013909

Cited by 2 publications

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Vo et. al [4] used CeyMo Road Marking dataset for detection and identification road markings. They have experimented and evaluated performance of the different object detection models and algorithms.…”

Section: Literature Surveymentioning

confidence: 99%

SAR Image based Target Detection & Identification using Deep Learning Model YOLOv7

2023

5th ISSE National Conference (INAC-05) on Systems Approach for Self-Reliance in Advanced Technologies

View full text Add to dashboard Cite

In this paper we have implemented a Real Time ship Target Detection on SAR Imagery Using deep learning model. Ship detection is crucial to neutralize marine targets during combat scenario. SAR images operate in allweather conditions and all-day and independent of range. SAR imagery-based surveillance technology is particularly well suited for marine surveillance. An Open-source SAR-ship dataset (HRSID) along with custom dataset is used for detection and identification. The existing AI based methods of detection (two stage object detectors) have proposed regions followed by extraction of features. The speed of the process for these methods is slow for real-time applications. Hence one stage detector, YOLOv7 with 415 convolution layers is used for the application which will overcome constraints of the previous methods. Mean Average Precision (mAP) of 78% @ 0.5 IOU is achieved for test HRSID dataset and custom dataset. NVIDIA GPU (RTX A4000) hardware is used for the training. The inference time of 8.2 milliseconds is achieved on NVIDIA RTX A4000 GPU. Hence YOLOv7 is suitable for real time detection of military targets on SAR images.

show abstract

Section: Literature Surveymentioning

confidence: 99%