Desarrollo de una aplicación móvil de reconocimiento de suelos usando deep learning.

Abstract

The inclusion of artificial intelligence in technological projects has increased in recent years, this is because there is great potential in this type of technology, because with artificial intelligence several activities can be carried out, such as classifying images for detect different objects and in this way be able to identify which object is the one that is shown in a certain window of an application. Artificial intelligence comprises several aspects such as the use of neural networks to carry out the training processes of a model that will be created through the use of special programming languages for said approach, in addition, it is necessary to previously have image data to classify for said training. Currently, you can find in different articles in scientific journals many projects that are related to the implementation of mobile applications that include in their functions the use of artificial intelligence to carry out certain processes that require image analysis to detect, for example, caries or bacterial plaque in the photos of the mouths of a group of patients selected as data for the neural network that is being used or for the artificial vision that is being used. Knowing the above, a soil recognition mobile application was developed using Deep Learning, which through a training model previously created with convolutional neural networks is consumed to make the respective predictions of the images captured from the cell phone or photos from the device's gallery. , indicating its probability and soil name. The development of this mobile application was carried out with the Xamarin Framework, which is a multiplatform environment that works with the programming language C# and XAML. Logic and Design respectively, in addition certain libraries were used to work with TensorFlow, which is another Framework used to create and use neural networks. The XP methodology was used for the construction of the project due to its agile approach and for giving more priority to development time than to communication with stakeholders. To use the application, the user must choose between two options to proceed with the analysis of the image or photo. In the initial window a menu will appear in which you can choose between selecting an image from the gallery or taking a photo, followed by a new window called analysis in which you can see the image obtained from the previously selected option, In addition, a message will appear indicating the name of the soil detected with its probability of certainty and a list of the other soils with their respective probabilities of certainty ordered in descending order. To evaluate the mobile application, the ISO-IEC 25010 quality standard was used, because this standard focuses on applying several quality evaluation metrics to the final product, that is, to the developed mobile application, in addition to these metrics tests are also used. units that confirm compliance with the user stories outlined in this report.