Object Detection using Convolutional Neural Network Transfer Learning

Authors: E L Seetam, F E Onuodu

Country: Nigeria

Full-text Research PDF File:   View   |   Download

Abstract: Any machine learning algorithm's ability to extract salient (relevant) characteristics is critical to its success. Traditional machine learning methods rely on domain expert-generated input features or computational feature extraction techniques. A convolutional neural network (CNN) is a type of artificial intelligence inspired by how the human brain's visual cortex functions when it comes to object detection. Because CNN requires a large number of neurons and layers to train data, it is not ideal for small datasets. Obtaining and storing a huge data collection for a scratch program is a challenge. These issues can be solved by using transfer training using a pre-trained data set. This is a dimensionality reduction approach used in deep learning analysis to lower the number of hidden layers and construct neural network applications on tiny data sets with high gain and little information loss. Using transfer learning to retrain a convolutional neural network to categorize a fresh batch of photos, this research investigates visual properties and isolates those that unify the digital image. The developed model satisfied 97% MSE.

Keywords: CNN, Transfer Leaning, Pre-trained Image, Computer Vision

Paper Id: 1371

Published On: 2022-06-12

Published In: Volume 10, Issue 3, May-June 2022

Cite This: Object Detection using Convolutional Neural Network Transfer Learning - E L Seetam, F E Onuodu - IJIRMPS Volume 10, Issue 3, May-June 2022.

Share this