Computer vision application proposal for smart inventory systems in convenience store reach-in refrigerators
Fecha
2022-08Autor
Ortega-Tapetillo, Juan C.
Contreras-López, Rodrigo
Metadatos
Mostrar el registro completo del ítemDescripción
Inventory systems in reach-in refrigerators employ manual or smart inventory outdated methods, although efficient, new methods like computer vision could render better results in less time, with less human intervention. The objective of this work proposes a computer vision system to acquire an inventory of products placed in reach-in convenience store refrigerators. A comparative of different computer vision object recognition models was performed to select the most appropriate model for the application. Then, based on the model characteristics and the application requirements, a YOLOv4 object recognition model was selected. Along with a 2-dimension camera positioning rig to capture a live video feed of the products to count for the inventory. Future works could include a real size prototype and further development into a commercial product.ITESO, A. C.