Izvorni znanstveni članak

https://doi.org/10.17559/TV-20230709000793

EODM: On Developing Enhanced Object Detection Model using Fast Region-based Convolution Neural Networks (FRCNN)

Anuradha B. ; Department of Computer Science & Engineering, SNS College of Engineering, Coimbatore-641107, Tamil Nadu, India
Karthik S. ; Department of Computer Science & Engineering, SNS College of Technology, Coimbatore-641035, Tamil Nadu, India *
Mythili S. ; Department of Computer Science & Engineering, SNS College of Technology, Coimbatore-641035, Tamil Nadu, India
Kavitha M. S. ; Department of Computer Science & Engineering, SNS College of Technology, Coimbatore-641035, Tamil Nadu, India

* Dopisni autor.

Sažetak

In present scenario, in machine learning technology, computer vision technology and image processing have attained a massive growth. Amongst many branches of image processing and classification, Object Detection (OD) is the major research domain. In several domains such as face detection, self-driving cars, pedestrian detection, and security surveillance systems, object detection (OD) and classification have experienced a significant surge in popularity in recent years. The conventional techniques for object detection, such as background removal, Gaussian Mixture Model (GMM), and Support Vector Machine (SVM), exhibit limitations such as object overlap, distortion caused by environmental factors including smoke, fog, and varying lighting conditions.Though there are several methods developed for OD, the respective field still stumbles upon many confrontations at the real-time implementations. Detecting objects from the undefined background is the major problem to be considered. Hence, machine learning techniques are incorporated for detecting the objects accurately, when the Neural Networks are effectively trained. With that note, this paper develops a new model, called Enhanced Object Detection Model using Fast Region-based Convolution Neural Networks (FRCNN). For producing appropriate results, sensitivity Measurement is carried out based on brightness, saturation, contrast, Gaussian blur, Gaussian Noise and sharpness. Following this, FRCNN is trained for OD and the results are obtained. The model evaluations are carried out based on some evaluation factors with the acquired dataset images. The obtained results are compared with CNN, YOLO. The result shows that the model exemplifies the other compared works in terms of efficiency and accuracy.

Ključne riječi

accuracy; computer vision; CNN; image processing; machine learning; object detection; sensitivity

Hrčak ID:

314848

URI

https://hrcak.srce.hr/314848

Datum izdavanja:

29.2.2024.

Posjeta: 866 *