A
Alexander Freytag
Researcher at Carl Zeiss AG
Publications - 37
Citations - 1053
Alexander Freytag is an academic researcher from Carl Zeiss AG. The author has contributed to research in topics: Active learning (machine learning) & Novelty detection. The author has an hindex of 15, co-authored 32 publications receiving 845 citations. Previous affiliations of Alexander Freytag include University of Jena & Schiller International University.
Papers
More filters
Book ChapterDOI
Selecting Influential Examples: Active Learning with Expected Model Output Changes
TL;DR: The key idea of this approach is to measure the expected change of model outputs, a concept that generalizes previous methods based on expected model change and incorporates the underlying data distribution.
Proceedings ArticleDOI
Kernel Null Space Methods for Novelty Detection
TL;DR: This work presents how to apply a null space method for novelty detection, which maps all training samples of one class to a single point, which outperforms all other methods for multi-class novelty detection.
Proceedings ArticleDOI
Nonparametric Part Transfer for Fine-Grained Recognition
TL;DR: An approach for fine-grained recognition based on a new part detection method which transfers part constellations from objects with similar global shapes is presented and the importance of carefully designed visual extraction strategies, including combination of complementary feature types and iterative image segmentation, is shown.
Book ChapterDOI
Fine-Tuning Deep Neural Networks in Continuous Learning Scenarios
TL;DR: This paper applies the classical concept of fine-tuning deep neural networks to scenarios where data from known or completely new classes is continuously added, and empirically analyzes how computational burdens of training can be further reduced.
Book ChapterDOI
Chimpanzee Faces in the Wild: Log-Euclidean CNNs for Predicting Identities and Attributes of Primates
TL;DR: This paper builds on convolutional neural networks, which lead to significantly superior results compared with previous state-of-the-art on hand-crafted recognition pipelines, and shows how to further increase discrimination abilities of CNN activations by the Log-Euclidean framework on top of bilinear pooling.