WiDS Villach 2024 | Isabell Dicillia-Kovatsch & Corinna Kofler

The WiDS Villach 2024 talk given Isabell Dicillia-Kovatsch and Corinna Kofler focused on computer vision in the context of semiconductor manufacturing. Both speakers work in the field, focusing on defect density inspection using computer vision techniques. They discussed the significance of early defect detection and the application of machine learning models for defect classification.

Highlights:

🖼️Computer vision techniques such as image classification, object detection, and instance segmentation are used in diverse fields like self-driving cars, healthcare, surveillance systems, and manufacturing.

📊Data-centric approaches for model development work well. For these approaches' data preparation, model training, and validation are vital.

🧠To use of convolutional neural networks (CNNs) for image classification, along with transfer learning and hyperparameter tuning, data challenges need to be solved. E.g. class imbalance in the dataset. Techniques like resampling and data augmentation were discussed as methods to mitigate data imbalance and enhance model generalization.

🚀The importance of model monitoring and the challenges faced during deployment were also highlighted.

🌍Real-World Data Set Sharing: a subset of used dataset is publicly available at Zenodo and can be used for collaboration and further research in the field.

Overall, the talk provided valuable insights into the application of computer vision in semiconductor defect detection, emphasizing the collaborative and iterative nature of model development and the importance of real-world data sharing for advancing research in the field. If you need further details or specific information from the talk, feel free to ask!

Video credits: MC Digitalproduktions GmbH