Control de merma para la sección de productos de casa inteligente en una compañía de retail, basado en algoritmos de Machine Learning de última generación (AutoML y YOLO) integrando hardware de inferencia neuronal
| dc.contributor.advisor | Puentes Morales, Carlos Alberto | |
| dc.contributor.author | Castellanos Chacon, Sergio Camilo | |
| dc.date.accessioned | 2026-02-12T14:32:00Z | |
| dc.date.issued | 2023-09 | |
| dc.description.abstract | Este artículo presenta el desarrollo e implementación de una solución basada en algoritmos de aprendizaje automático de última generación y hardware de inferencia neuronal para la detección y prevención de la merma en tiempo real en una compañía de retail. La merma, entendida como la pérdida de inventario por factores como el hurto, el vencimiento o la manipulación inadecuada de productos, representa un problema significativo para el sector retail en Colombia. La metodología propuesta incluyó la recolección de 530 imágenes de productos de casa inteligente, su depuración mediante criterios de inclusión y exclusión, y la construcción de un conjunto de datos final compuesto por 479 imágenes etiquetadas manualmente. Posteriormente, se aplicaron técnicas de preprocesamiento y aumento de datos para mejorar la generalización de los modelos. Se entrenaron y evaluaron modelos de detección de objetos basados en AutoML, YOLOv5 y YOLOv8, utilizando métricas como mean Average Precision (mAP), precisión y recall. Los resultados mostraron que el modelo AutoML alcanzó el mejor desempeño, con un mAP del 94,1 %, una precisión del 91,4 % y un recall del 92,4 %. Dicho modelo fue integrado con una cámara OAK-1 para su despliegue en un entorno real de tienda, permitiendo la generación de inferencias en tiempo real y su posterior comparación con los registros de ventas. La solución implementada demostró ser efectiva para identificar posibles casos de merma, contribuyendo a mejorar el control de inventario y la eficiencia operativa en el retail. | |
| dc.description.abstractenglish | This article presents the development and implementation of a solution based on state-of-the-art machine learning algorithms and neural inference hardware for the real-time detection and prevention of retail shrinkage. Shrinkage, defined as inventory loss caused by factors such as theft, expiration, or improper handling of products, represents a significant challenge for the retail sector in Colombia. The proposed methodology involved the collection of 530 images of smart home products, their refinement through inclusion and exclusion criteria, and the construction of a final dataset composed of 479 manually labeled images. Subsequently, image preprocessing and data augmentation techniques were applied to improve model generalization. Object detection models based on AutoML, YOLOv5, and YOLOv8 were trained and evaluated using metrics such as mean Average Precision (mAP), precision, and recall. The results showed that the AutoML model achieved the best performance, with an mAP of 94.1%, a precision of 91.4%, and a recall of 92.4%. This model was integrated with an OAK-1 camera and deployed in a real retail environment, enabling real-time inference generation and comparison with sales records. The implemented solution proved to be effective in identifying potential shrinkage events, thereby improving inventory control and operational efficiency in retail settings. | |
| dc.identifier.uri | https://hdl.handle.net/20.500.12495/19270 | |
| dc.language.iso | es | |
| dc.relation.references | Asad, M. (2022). Pothole detection using deep learning: A real-time and AI-on-the-edge perspective. Applied Computational Intelligence. | |
| dc.relation.references | Chun, Y. (2020). Novel clothing attribute representation network-based self-attention mechanism. IEEE. | |
| dc.relation.references | Cinco Días. (2019). Las cámaras japonesas que pueden detectar robos antes de que sucedan. https://cincodias.elpais.com/cincodias/2019/03/04/companias/1551722564_039430.html | |
| dc.relation.references | Córdova, D. (2023). A comprehensive review of YOLO: From YOLOv1 to YOLOv8 and beyond. https://arxiv.org/abs/2304.00501 | |
| dc.relation.references | Di, C. (2023). Machine learning and computer vision for the automation of processes in advanced logistics: The integrated logistic platform. Procedia Computer Science. | |
| dc.relation.references | El Colombiano. (2020). Las pérdidas de inventarios en los supermercados crecieron 7 % en 2019. https://www.elcolombiano.com/negocios/empresas/perdidas-y-robos-de-supermercados-llegaron-a-589747-millones-de-pesos-en-2019-fenalco-DJ13923563 | |
| dc.relation.references | Fenalco. (2020). Censo nacional de merma. https://fenalcovalle.com/wp-content/uploads/2021/03/Presentacion-Gondola-2020.pdf | |
| dc.relation.references | Gaudenz, B. (2023). The 10 top applications of computer vision in retail in 2023. https://viso.ai/applications/computer-vision-in-retail/ | |
| dc.relation.references | González, N. (2022). El uso de tecnología e inteligencia artificial en el sector del retail aumenta hasta 40 % las ventas. https://www.larepublica.co/empresas/el-uso-de-tecnologia-e-inteligencia-artificial-en-el-retail-aumenta-hasta-40-las-ventas-3433686 | |
| dc.relation.references | Google Cloud. (2023a). Beginner’s guide to AutoML. https://cloud.google.com/vertex-ai/docs/beginner/beginners-guide | |
| dc.relation.references | Google Cloud. (2023b). AutoML Vision documentation. https://cloud.google.com/vision/automl/docs | |
| dc.relation.references | Hui, J. (2018). Real-time object detection with YOLO, YOLOv2, and YOLOv3. https://jonathan-hui.medium.com/real-time-object-detection-with-yolo-yolov2-28b1b93e2088 | |
| dc.relation.references | Isahit. (s. f.). 10 computer vision applications in retail. https://www.isahit.com/blog/10-computer-vision-applications-in-retail | |
| dc.relation.references | Jeon, J.-Y. (2022). A retail object classification method using multiple cameras for vision-based unmanned kiosks. IEEE Access. | |
| dc.relation.references | Ji, Q. (2020). Probabilistic graphical models for computer vision. Academic Press. | |
| dc.relation.references | Joshi, N. (2022). How computer vision is transforming the retail industry. https://www.allerin.com/blog/how-computer-vision-is-transforming-the-retail-industry | |
| dc.relation.references | Kaur, J. (2022). Tools, techniques, datasets and application areas for object detection in an image: A review2eview. Multimedia Tools and Applications. | |
| dc.relation.references | Kundu, R. (2023). YOLO: Algorithm for object detection explained. https://www.v7labs.com/blog/yolo-object-detection | |
| dc.relation.references | Luxonis. (2020). OAK-1 camera. https://shop.luxonis.com/collections/oak-cameras-1/products/oak-1 | |
| dc.relation.references | Ma, R. (2021). Human motion gesture recognition based on computer vision. Complexity. | |
| dc.relation.references | Matallanas, E. (2022). Aplicaciones de la inteligencia artificial en el sector retail. https://www.plainconcepts.com/es/soluciones-inteligencia-artificial-retail/ | |
| dc.relation.references | Mohamad. (2016). SSE – Suma de cuadrados de los residuales o errores. https://support.numxl.com/hc/es/articles/215653803 | |
| dc.relation.references | Nixon, M. (2020). Feature extraction and image processing for computer vision. Academic Press. | |
| dc.relation.references | Odemakinde, E. (2023). The 27 most popular computer vision companies and startups in 2023. https://viso.ai/computer-vision/most-popular-computer-vision-companies-and-startups/ | |
| dc.relation.references | Ou, B. (2022). Analysis of people flow image detection system based on computer vision sensor. Journal of Sensors. | |
| dc.relation.references | Pérez, L. (2018). ¿Qué es la merma en el retail? https://es.linkedin.com/pulse/que-es-la-merma-en-el-retail-martha-elia-p%C3%A9rez-l%C3%B3pez | |
| dc.relation.references | Portafolio. (2022). Los efectos en los supermercados por el aumento de los robos. https://www.portafolio.co/economia/finanzas/los-efectos-en-los-supermercados-por-el-aumento-de-los-robos-573231 | |
| dc.relation.references | Senarath, S. (2022). Customer gaze estimation in retail using deep learning. IEEE Sensors Journal. | |
| dc.relation.references | Sensormatic. (s. f.). Computer vision and the future of retail. https://d5544430a84c15063ea9-24a29c251add4cb0f3d45e39c18c202f.ssl.cf1.rackcdn.com/Sensormatic_Computer%20Vision_Final.pdf | |
| dc.relation.references | Torres. (s. f.). Data augmentation y transfer learning. https://torres.ai/data-augmentation-y-transfer-learning-en-keras-tensorflow/ | |
| dc.relation.references | Ultralytics. (2023a). Ultralytics YOLOv8 documentation. https://docs.ultralytics.com/ | |
| dc.relation.references | Ultralytics. (2023b). Ultralytics YOLO GitHub repository. https://github.com/ultralytics/ultralytics | |
| dc.relation.references | Ultralytics. (2023c). YOLO version comparison. https://github.com/ultralytics/ultralytics | |
| dc.rights | Attribution 4.0 International | en |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | |
| dc.subject | Merma | |
| dc.subject | Retail | |
| dc.subject | Visión por computador | |
| dc.subject | Aprendizaje automático | |
| dc.subject | AutoML | |
| dc.subject | YOLO | |
| dc.subject | Inferencia neuronal | |
| dc.subject.keywords | Shrinkage | |
| dc.subject.keywords | Retail | |
| dc.subject.keywords | Computer vision | |
| dc.subject.keywords | Machine learning | |
| dc.subject.keywords | AutoML | |
| dc.subject.keywords | YOLO | |
| dc.subject.keywords | neural inference | |
| dc.title | Control de merma para la sección de productos de casa inteligente en una compañía de retail, basado en algoritmos de Machine Learning de última generación (AutoML y YOLO) integrando hardware de inferencia neuronal | |
| dc.title.translated | Shrinkage control for the smart home products section in a retail company based on stateof-the-art Machine Learning Algorithms (AutoML and YOLO) integrated with neural inference hardware |
Archivos
Bloque original
1 - 1 de 1
Cargando...
- Nombre:
- Trabajo de grado.pdf
- Tamaño:
- 1.43 MB
- Formato:
- Adobe Portable Document Format
Bloque de licencias
1 - 3 de 3
Cargando...
- Nombre:
- license.txt
- Tamaño:
- 2.06 KB
- Formato:
- Item-specific license agreed upon to submission
- Descripción:
Cargando...
- Nombre:
- Carta de autorizacion.pdf
- Tamaño:
- 131.62 KB
- Formato:
- Adobe Portable Document Format
- Descripción:
Cargando...
- Nombre:
- Anexo 1 Acta de aprobacion.pdf
- Tamaño:
- 495.47 KB
- Formato:
- Adobe Portable Document Format
- Descripción:
