Fraunhofer-Institut für Produktionsanlagen und Konstruktionstechnik IPK

Filters
2
1 1
2
Reset filters

Settings
Now showing 1 - 2 of 2
  • Publication
    With synthetic data towards part recognition generalized beyond the training instances
    ( 2024)
    Koch, Paul
    ;
    Schlüter, Marian
    ;
    Krüger, Jörg
    In this work we investigate the effect of using synthetic data, generated in a simulation, in order to pre-train an AI-based image classification for industrial components. After pre-training we use real camera-captured training images to fine-tune the AI with the aim to close the Sim2Real domain gap. We compare our approach to purely using real training images of a single candidate object instance. In an exemplary case study for screw recognition, we found that a given AI classification algorithm dropped its recognition rate from 99.8% to 88.5% when testing the algorithm with known and unknown screw instances of the learned object classes, respectively. Employing our pre-training method on the basis of synthetic data, the drop in recognition rate is decreased from 99% to 96.95%. Thus, our proposed method has only a relative drop of 2.05% when shifting towards a generalized domain (including unknown part instances), while a compared approach on the basis of real camera-captured data showed a drop of 11.3%.
  • Publication
    Classification of Similar Objects of Different Sizes Using a Reference Object by Means of Convolutional Neural Networks
    ( 2019)
    Lehr, Jan
    ;
    Schlüter, Marian
    ;
    Krüger, Jörg
    Part identification is relevant in many industrial applications, either for direct recognition of components or assemblies, either as a fully automated process or as an assistance system. Convolutional Neural Networks (CNNs) have proven their worth in image processing, especially in classification tasks. It therefore makes sense to use them for industrial applications. There are major problems with parts that look very similar and can only be identified by their size. In this paper we have considered a subset of screws that all conform to the same norm but are of different sizes. The implicit learning of the screw size is only possible if the images are taken in a fixed distance setup and larger screws are shown larger on the images. In this paper we show that CNNs are able to implicitly measure target objects with the help of reference objects and thus to integrate the object size into the learning process.