Vlado Menkovski bio photo

Vlado Menkovski

Assistant Professor at Eindhoven University of Technology.

Email Google Scholar

TLDR; Computer vision algorithms trained on very few samples, understand the function and anomalies of micro-electronic devices in production environments (application to defect detection, metrology, alignments, and vision-driven control)

Project Description: A key challenge for using deep learning and other contemporary data-driven techniques in advanced micro-manufacturing applications is the scarcity of labeled data needed for training. In many tasks, such as defect inspection, training samples are inherently rare. In other cases, frequent changes in the layout of the manufactured devices makes the task of collecting and labeling large amounts of data costly. The goal of this project is to develop vision algorithms that can learn from less data with applications in inspection, alignment, metrology, and micro-manufacturing process control. While representative samples of anomalies are rare, a large amount of un-labeled data is continuously recorded from vision cameras and other sensors. Additionally, large labeled datasets of different types of devices, than the ones being produced, are often at hands. Such datasets can be exploited to develop techniques generalizing component recognition and function reasoning to new unseen assemblies. In this project we focus particularly on use-cases where a limited number of samples are labeled on the production floor, in a short time and with low quality annotation. The goal is to develop methods that can provide good detection, semantic segmentation, and anomaly reasoning results from such noisy and limited labeling. The basis of such methods may reside in techniques such as: semi-supervised, weakly-supervised, and distantly-supervised learning. The latter approach is possibly the most relevant to the considered use-cases and applications.

Company Description: The ASM Center of Competency, located in the Netherlands, is responsible for identifying new high-tech innovation opportunities for the full global ASM Pacific Technology product portfolio. The equipment portfolio of ASM Pacific Technology covers the complete back-end section of the semiconductor equipment market, and offers comprehensive assembly and advanced packaging solutions for its customers in the fields of optoelectronics, electronics, solar energy, automotive and other segments.