Artificial Intelligence and Communication Technologies
Image and its Coordinates Detection in Convolution Neural Network Using YOLO Framework
Authors: Abhilasha Gupta, Krishna Joshi and Umesh Diwedi
Publishing Date: 31-03-2023
ISBN: 978-81-955020-5-9
Abstract
Computer vision is a field that deals with high level under- standing from digital Images. It provides new directions for making machines attentive and responsive to man. In this paper we are propos- ing an approach that is used for image detection. Face detection is a major problem in this area and many dataset are created for the pur- pose of computer vision. We have various deep learning models like convolutional neural network, recurrent neural network etc. But among all, deep convolutional neural networks are the best model for finding patterns from images. In the same direction we designed a model which receives its input from three fully connected layer that predict the coor- dinates and probability, and fully connected layer gets its input from Pooling layer that would scale back number of parameters when pic- tures are over large. To detect the image features and its coordinates we are using bounding boxes. Image detection is performing with the help of convolution neural network that is getting its input from three fully connected layers and fully connected layer is getting its input from Pooling layer. We used Relu layer to provide input to pooling layer. In this way images features are classified and the image is predicted on the basis of this model. So our output layer predicts both class proba- bility and coordinates of bounding boxes. The work is tested with the help of two experiment setup on two different machines. Results gave remarkable achievement and future directions.
Keywords
Image detection, Convolution Layer, Relu Layer.
Cite as
Abhilasha Gupta, Krishna Joshi and Umesh Diwedi, "Image and its Coordinates Detection in Convolution Neural Network Using YOLO Framework", In: Saroj Hiranwal and Garima Mathur (eds), Artificial Intelligence and Communication Technologies, SCRS, India, 2023, pp. 905-909. https://doi.org/10.52458/978-81-955020-5-9-86