Automation of the Labeling Process Using an Image Classification Model Using Convolutional Neural Networks

Emerging technologies enable the refocusing of communication and interaction strategies with clients and users through significant innovations, such as the use of 360 degrees videos and immersive Virtual Reality (VR) in tourism and hotel promotion. The aim of this work is to leverage these technologies to optimize the creation of 360 degrees experiences through process automation focused on the classification of images that will compose such experiences. In our proposal, we designed a Convolutional Neural Network (CNN), whose essential functions are feature extraction and image classification and output processes, as these will be used for the composition of virtual tours. The feature extraction stage consists of several hidden layers, such as the convolution layer, the Rectified Linear Unit (ReLU) activation function, and the pooling layer. Subsequently, training and testing are conducted to ensure that the labeling process of 360 degrees videos is automated by the virtual tour viewer prototype, optimizing a process traditionally performed manually. The model's functionalities and test results were satisfactory, achieving 95.09% accuracy, surpassing the success indicators for such a model. Finally, conclusions and recommendations for future work are established.