Automatic localization of casting defects with convolutional neural networks

Ferguson, Max; Ak, Ronay; Lee, Yung-Tsun Tina; Law, Kincho H.

doi:10.1109/bigdata.2017.8258115

Cited by 156 publications

(65 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The size of the images in the dataset ranges from 256 × 256 pixels to 768 × 572 pixels. To ensure the results are consistent with previous work, the training and testing data is divided in the same way as described in [42].…”

Section: Implementantion Details and Experimental Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Detection and Segmentation of Manufacturing Defects with Convolutional Neural Networks and Transfer Learning

Ferguson

Lee

et al. 2018

Smart and Sustainable Manufacturing Systems

Self Cite

187

View full text Add to dashboard Cite

Quality control is a fundamental component of many manufacturing processes, especially those involving casting or welding. However, manual quality control procedures are often time-consuming and error-prone. In order to meet the growing demand for high-quality products, the use of intelligent visual inspection systems is becoming essential in production lines. Recently, Convolutional Neural Networks (CNNs) have shown outstanding performance in both image classification and localization tasks. In this article, a system is proposed for the identification of casting defects in X-ray images, based on the Mask Region-based CNN architecture. The proposed defect detection system simultaneously performs defect detection and segmentation on input images, making it suitable for a range of defect detection tasks. It is shown that training the network to simultaneously perform defect detection and defect instance segmentation, results in a higher defect detection accuracy than training on defect detection alone. Transfer learning is leveraged to reduce the training data demands and increase the prediction accuracy of the trained model. More specifically, the model is first trained with two large openly-available image datasets before finetuning on a relatively small metal casting X-ray dataset. The accuracy of the trained model exceeds state-of-the art performance on the GRIMA database of X-ray images (GDXray) Castings dataset and is fast enough to be used in a production setting. The system also performs well on the GDXray Welds dataset. A number of in-depth studies are conducted to explore how transfer learning, multi-task learning, and multi-class learning influence the performance of the trained system.

show abstract

Section: Implementantion Details and Experimental Resultsmentioning

confidence: 99%

“…Comparison of the accuracy and performance of each model on the defect detection task. Results are compared to the previous state-of-the-art results, presented in[42]. augmentation techniques on test accuracy.…”

mentioning

confidence: 99%

Detection and Segmentation of Manufacturing Defects with Convolutional Neural Networks and Transfer Learning

Ferguson

Lee

et al. 2018

Smart and Sustainable Manufacturing Systems

Self Cite

187

View full text Add to dashboard Cite

show abstract

“…Actualmente, se han desarrollado sistemas de visión computacional que permiten examinar automáticamente la apariencia visual de los productos en las líneas de producción. Muchos de estos sistemas usan estrategias de aprendizaje profundo (Cao et al, 2018(Cao et al, , 2016Park et al, 2016;Chen et al, 2018;Shi et al, 2017;Ferguson et al, 2017). Por otra parte, se han desarrollado sistemas que se enfocan en la monitorización de las condiciones de las herramientas utilizadas durante la producción, con el fin de identificar de forma temprana los fallos en el sistema (García-Ordás et al, 2017;Fernández-Robles et al, 2017a,b).…”

Section: Inspección De Calidadunclassified

“…Además, la localización previa es fundamen-tal en el análisis automático de imágenes, cuyo objetivo puede ser la segmentación, que consiste en separar los objetos de interés del fondo, en el reconocimiento o recuperación de objetos (Saikia et al, 2017;García-Olalla et al, 2018), o en el análisis de la relación espacial entre los objetos contenidos en una imagen (Lampert et al, 2008). Actualmente, hay una gran cantidad de aplicaciones que requieren una localización precisa de los objetos, como es el caso de la necesidad que tienen los vehículos autónomos de localizar peatones (Dollár et al, 2009;Li et al, 2018;Du et al, 2017;Brazil et al, 2017;Wang et al, 2018) u obstáculos (Shah et al, 2018;Yi et al, 2018;Garnett et al, 2017;Sepúlveda et al, 2017), la localización de vehículos en sistemas de control de tráfico, se encuentren o no en imágenesáereas (Zhong et al, 2017;Ammour et al, 2017;Tang et al, 2017;Xu et al, 2017a;Lee et al, 2017), la localización de lesiones o anormalidades en tejidos que usan los sistemas de diagnóstico clínico asistido por computador (He et al, 2018;Ma et al, 2017;Jiamin et al, 2017;Sa et al, 2017;Heo et al, 2017), la detección de objetos para el control de calidad que requieren los sistemas de inspección visual (Cao et al, 2018;Chen et al, 2018;Shi et al, 2017;Ferguson et al, 2017), o la localización de obstáculos que tienen que realizar los sistemas de navegación de robots (Lee et al, 2015;Luo et al, 2017), entre otras. Sin embargo, la correcta localización de objetos es difícil debido a múltiples factores, entre los que destacan la falta de calidad de la imagen, condiciones de iluminación cambiantes, objetos con forma no rígida o los cambios en la apariencia de los objetos a localizar (Felzenszwalb et al, 2010;…”

Section: Introductionunclassified

Una Revisión Sistemática de Métodos para Localizar Automáticamente Objetos en Imágenes

Chaves

Saikia

Fernández-Robles

et al. 2018

Rev. iberoam. autom. inform. ind.

View full text Add to dashboard Cite

<p>Actualmente, muchas aplicaciones requieren localizar de forma precisa los objetos que aparecen en una imagen, para su posterior procesamiento. Este es el caso de la inspección visual en la industria, los sistemas de diagnóstico clínico asistido por computador, la detección de obstáculos en vehículos o en robots, entre otros. Sin embargo, diversos factores como la calidad de la imagen y la apariencia de los objetos a detectar, dificultan la localización automática. En este artículo realizamos una revisión sistemática de los principales métodos utilizados para localizar objetos, considerando desde los métodos basados en ventanas deslizantes, como el detector propuesto por Viola y Jones, hasta los métodos actuales que usan redes de aprendizaje profundo, tales como Faster-RCNNo Mask-RCNN. Para cada propuesta, describimos los detalles relevantes, considerando sus ventajas y desventajas, así como sus aplicaciones en diversas áreas. El artículo pretende proporcionar una revisión ordenada y condensada del estado del arte de estas técnicas, su utilidad y sus implementaciones a fin de facilitar su conocimiento y uso por cualquier investigador que requiera localizar objetos en imágenes digitales. Concluimos este trabajo resumiendo las ideas presentadas y discutiendo líneas de trabajo futuro.</p>

show abstract

“…To obtain effective 3D defects shapes, Zhao et al [19] combined line array CCD and area array CCD imaging methods and devised the informative image scanning method. As a fast developing subfield of machine learning, multilayer perceptron convolutional neural networks combined with deep learning(CNN-DL) strategies in MV inspection field have shown state-of-the-art performance [20]. CNN-DL methods do not require laborious hand-craft features for classifier design [21] and as a branch of ANN, they make the complex function approximation feasible by learning a deep nonlinear network.…”

Section: Introductionmentioning

confidence: 99%

A Deep-Learning-based 3D Defect Quantitative Inspection System in CC Products Surface

Zhao

Zhang

et al. 2020

Sensors

View full text Add to dashboard Cite

To create an intelligent surface region of interests (ROI) 3D quantitative inspection strategy a reality in the continuous casting (CC) production line, an improved 3D laser image scanning system (3D-LDS) was established based on binocular imaging and deep-learning techniques. In 3D-LDS, firstly, to meet the requirements of the industrial application, the CCD laser image scanning method was optimized in high-temperature experiments and secondly, we proposed a novel region proposal method based on 3D ROI initial depth location for effectively suppressing redundant candidate bounding boxes generated by pseudo-defects in a real-time inspection process. Thirdly, a novel two-step defects inspection strategy was presented by devising a fusion deep CNN model which combined fully connected networks (for defects classification/recognition) and fully convolutional networks (for defects delineation). The 3D-LDS’ dichotomous inspection method of defects classification and delineation processes are helpful in understanding and addressing challenges for defects inspection in CC product surfaces. The applicability of the presented methods is mainly tied to the surface quality inspection for slab, strip and billet products.

show abstract

Automatic localization of casting defects with convolutional neural networks

Cited by 156 publications

References 18 publications

Detection and Segmentation of Manufacturing Defects with Convolutional Neural Networks and Transfer Learning

Detection and Segmentation of Manufacturing Defects with Convolutional Neural Networks and Transfer Learning

Una Revisión Sistemática de Métodos para Localizar Automáticamente Objetos en Imágenes

A Deep-Learning-based 3D Defect Quantitative Inspection System in CC Products Surface

Contact Info

Product

Resources

About