Examinando por Autor "Soto Ríos, Juan Nicolas"
Mostrando 1 - 1 de 1
Resultados por página
Opciones de ordenación
- ÍtemClasificador de latas sin barniz interno dentro del proceso de fabricación de envases de aluminio para bebidas(Fundación Universitaria Los Libertadores. Sede Bogotá., ) Soto Ríos, Juan Nicolas; Gonzalez Veloza, Jose John FredyThis article proposes the development of an aluminum can classifier without internal varnish of a production line, from the training of an existing model called MobileNet V2, using a specific data set. 1766 images of cans were collected from a production line, which included containers with varnish and without internal varnish. Images of cans with internal varnish were labeled with [OK] and those containing cans without varnish with [W]. The data was randomly partitioned into two groups and from a validation process, the proportionality of images with labels [W] and [OK] was ensured for both groups of the DataSet. Subsequently, the data preparation was carried out, converting the images to tensors, where the size of the images is changed to that required by the model (224x224). From algorithm training, an overall precision of 0.88 and a false negative ratio for the unvarnished cans category [W] of 0.1 were achieved. Therefore, the model manages to correctly classify 88% of the images, correctly identifying those that have varnish and those that do not. In this way, from the application of an image classification tool, driven by Deep Learning algorithms, it is possible to solve problems in the industrial area. In this specific case, the absence of internal varnish in aluminum cans, reducing failures in the process and ensuring the quality of the final product.