Authors:
Thiago Leonardo Maria
1
;
Saul Delabrida
2
;
1
and
Andrea Gomes Campos
1
;
2
Affiliations:
1
Graduate Program in Instrumentation, Control and Automation of Mining Processes (PROFICAM), Federal University of Ouro Preto (UFOP) and Vale Institute of Technology (ITV), Minas Gerais, Brazil
;
2
Department of Computing (DECOM), Federal University of Ouro Preto (UFOP), Ouro Preto, Brazil
Keyword(s):
Wagon, Object Detection, Synthetic Data.
Abstract:
Efficient wagon loading plays a crucial role in logistic efficiency and supplying essential raw materials to various industries. However, ensuring the cleanliness of the wagons before loading is a critical aspect of this process as it directly impacts the quality and integrity of the transported item. Early detection of objects inside empty wagons before loading is a key component in this logistic puzzle. This study proposes a computer vision approach for object detection in train wagons before loading and performs a comparison between two models: YOLO (You Only Look Once) and RT-DETR (Real-Time Detection Transformer), which are based on Convolutional Neural Networks (CNNs) and Transformers, respectively. Additionally, the research addresses the generation of synthetic data as a strategy for model training, using the \textit{Unity} platform to create virtual environments that simulate real conditions of wagon loading. Therefore, the findings highlight the potential of combining compu
ter vision and synthetic data to improve the safety, efficiency, and automation of train loading processes, offering valuable insights into the application of advanced vision models in industrial scenarios.
(More)