Object detection techniques : overview and performance comparison

Noman, Mohammed and Stankovic, Vladimir and Tawfik, Ayman (2019) Object detection techniques : overview and performance comparison. In: 2019 IEEE International Symposium on Signal Processing and Information Technology, 2019-12-10 - 2019-12-12, Ajman University Conference Center. (In Press)

[img]
Preview
Text (Noman-etal-ISSPIT2019-Object-detection-techniques-overview-and-performance-comparison)
Noman_etal_ISSPIT2019_Object_detection_techniques_overview_and_performance_comparison.pdf
Accepted Author Manuscript

Download (3MB)| Preview

    Abstract

    Object detection algorithms are improving by the minute. There are many common libraries or application program interface (APIs) to use. The most two common techniques ones are Microsoft Azure Cloud object detection and Google Tensorflow object detection. The first is an online-network based API, while the second is an offline-machine based API. Both have their advantages and disadvantages. A direct comparison between the most common object detection methods help in finding the best solution for advance system integration. This paper will discuss both methods and compare them in terms of accuracy, complexity and practicality. It will show advantages and also limitations of each method, and possibilities for improvement.