Logo des Repositoriums

Auflistung nach Schlagwort "Generalization"

1 - 3 von 3
  • Konferenzbeitrag
    Deep Learning Datasets Challenges For Semantic Segmentation - A Survey
    (INFORMATIK 2023 - Designing Futures: Zukünfte gestalten, 2023) Ponciano, Claire; Schaffert, Markus; Ponciano, Jean-Jacques
    This survey offers a comprehensive analysis of challenges encountered when employing large-scale datasets for deep learning-based semantic segmentation, an area with significant implica- tions for industries such as autonomous driving, precision agriculture, and medical imaging. Through a systematic review of 94 papers from Papers with Code, we identified 32 substantial challenges, which we categorized into six key areas: Data Quality and Quantity, Data Preprocessing, Resource Constraints, Data Management and Privacy, Generalization, and Data Compatibility. By identifying and explicating these challenges, our research provides a crucial reference point for future studies aiming to address these issues and enhance the performance of deep learning models for semantic segmentation. Future work will focus on leveraging AI and semantic technologies to provide solutions to these challenges.
  • Konferenzbeitrag
    Fake Face Detection Methods: Can They Be Generalized?
    (BIOSIG 2018 - Proceedings of the 17th International Conference of the Biometrics Special Interest Group, 2018) Khodabakhsh, Ali; Ramachandra, Raghavendra; Raja, Kiran; Wasnik, Pankaj; Busch, Christoph
    With advancements in technology, it is now possible to create representations of human faces in a seamless manner for fake media, leveraging the large-scale availability of videos. These fake faces can be used to conduct personation attacks on the targeted subjects. Availability of open source software and a variety of commercial applications provides an opportunity to generate fake videos of a particular target subject in a number of ways. In this article, we evaluate the generalizability of the fake face detection methods through a series of studies to benchmark the detection accuracy. To this extent, we have collected a new database of more than 53;000 images, from 150 videos, originating from multiple sources of digitally generated fakes including Computer Graphics Image (CGI) generation and many tampering based approaches. In addition, we have also included images (with more than 3;200) from the predominantly used Swap-Face application that is commonly available on smart-phones. Extensive experiments are carried out using both texture-based handcrafted detection methods and deep learning based detection methods to find the suitability of detection methods. Through the set of evaluation, we attempt to answer if the current fake face detection methods can be generalizable.
  • Zeitschriftenartikel
    Towards Confirmatory Process Discovery: Making Assertions About the Underlying System
    (Business & Information Systems Engineering: Vol. 61, No. 6, 2019) Janssenswillen, Gert; Depaire, Benoît
    The focus in the field of process mining, and process discovery in particular, has thus far been on exploring and describing event data by the means of models. Since the obtained models are often directly based on a sample of event data, the question whether they also apply to the real process typically remains unanswered. As the underlying process is unknown in real life, there is a need for unbiased estimators to assess the system-quality of a discovered model, and subsequently make assertions about the process. In this paper, an experiment is described and discussed to analyze whether existing fitness, precision and generalization metrics can be used as unbiased estimators of system fitness and system precision. The results show that important biases exist, which makes it currently nearly impossible to objectively measure the ability of a model to represent the system.