Tobias Schlagenhauf

Tobias Schlagenhauf, Yiwen Lin, Benjamin Noack

Ensembles of Convolutional neural networks have shown remarkable results in learning discriminative semantic features for image classification tasks. Though, the models in the ensemble often concentrate on similar regions in images. This work proposes a novel method that forces a set of base models to learn different features for a classification task. These models are combined in an ensemble to make a collective classification. The key finding is that by forcing the models to concentrate on different features, the classification accuracy is increased. To learn different feature concepts, a so-called feature distance loss is implemented on the feature maps. The experiments on benchmark convolutional neural networks (VGG16, ResNet, AlexNet), popular datasets (Cifar10, Cifar100, miniImageNet, NEU, BSD, TEX), and different training samples (3, 5, 10, 20, 50, 100 per class) show the effectiveness of the proposed feature loss. The proposed method outperforms classical ensemble versions of the base models. The Class Activation Maps explicitly prove the ability to learn different feature concepts. The code is available at: https://github.com/2Obe/Feature-Distance-Loss.git

Tobias Schlagenhauf, Markus Netzer, Jan Hillinger

This paper addresses the issue of autonomously detecting text on technical drawings. The detection of text on technical drawings is a critical step towards autonomous production machines, especially for brown-field processes, where no closed CAD-CAM solutions are available yet. Automating the process of reading and detecting text on technical drawings reduces the effort for handling inefficient media interruptions due to paper-based processes, which are often todays quasi-standard in brown-field processes. However, there are no reliable methods available yet to solve the issue of automatically detecting text on technical drawings. The unreliable detection of the contents on technical drawings using classical detection and object character recognition (OCR) tools is mainly due to the limited number of technical drawings and the captcha-like structure of the contents. Text is often combined with unknown symbols and interruptions by lines. Additionally, due to intellectual property rights and technical know-how issues, there are no out-of-the box training datasets available in the literature to train such models. This paper combines a domain knowledge-based generator to generate realistic technical drawings with a state-of-the-art object detection model to solve the issue of detecting text on technical drawings. The generator yields artificial technical drawings in a large variety and can be considered as a data augmentation generator. These artificial drawings are used for training, while the model is tested on real data. The authors show that artificially generated data of technical drawings improve the detection quality with an increasing number of drawings.

Tobias Schlagenhauf, Tim Scheurenbrand

A common challenge in real world classification scenarios with sequentially appending target domain data is insufficient training datasets during the training phase. Therefore, conventional deep learning and transfer learning classifiers are not applicable especially when individual classes are not represented or are severely underrepresented at the outset. In many technical domains, however, it is only the defect or worn reject classes that are insufficiently represented, while the non-defect class is often available from the beginning. The proposed classification approach addresses such conditions and is based on a CNN encoder. Following a contrastive learning approach, it is trained with a modified triplet loss function using two datasets: Besides the non-defective target domain class 1st dataset, a state-of-the-art labeled source domain dataset that contains highly related classes e.g., a related manufacturing error or wear defect but originates from a highly different domain e.g., different product, material, or appearance = 2nd dataset is utilized. The approach learns the classification features from the source domain dataset while at the same time learning the differences between the source and the target domain in a single training step, aiming to transfer the relevant features to the target domain. The classifier becomes sensitive to the classification features and by architecture robust against the highly domain-specific context. The approach is benchmarked in a technical and a non-technical domain and shows convincing classification results. In particular, it is shown that the domain generalization capabilities and classification results are improved by the proposed architecture, allowing for larger domain shifts between source and target domains.

Pascal Janetzky, Tobias Schlagenhauf, Stefan Feuerriegel

Continual learning (CL) aims to train models sequentially over multiple domains without forgetting previously learned knowledge. However, existing CL methods optimize for in-domain performance and are therefore prone to learning spurious, domain-specific cues (``shortcut learning''), which limits generalization to unseen domains after deployment. In this paper, we address this limitation through continual learning of domain-invariant representation. We introduce a broad class of CL methods that sequentially learn representations capturing invariant structures across domains. Our methods are motivated by the observation that such invariant structures often preserve the underlying causal mechanisms, which can reduce the risk of overfitting to domain-specific cues and thus offer better out-of-domain generalization. Our proposed CL methods combine replay-based training with a tailored sequential invariance alignment to learn -- and preserve -- invariant structures over time. We evaluate our methods under a deployment-oriented protocol that measures performance on unseen target domains. Across six benchmark and real-world datasets spanning vision, medicine, manufacturing, and ecology, our methods consistently outperform existing CL baselines in terms of generalization to unseen target domains. As an ablation, we further show that naïve extensions of sequential training with existing domain-invariant representation learning (DIRL) methods provide only limited benefits. To the best of our knowledge, this is the first work to develop domain-invariant representation methods for CL.

Zheng Wang, Kaixuan Zhang, Wanfang Chen et al.

Time series forecasting remains a critical challenge across numerous domains, yet the effectiveness of complex models often varies unpredictably across datasets. Recent studies highlight the surprising competitiveness of simple linear models, suggesting that their robustness and interpretability warrant deeper theoretical investigation. This paper presents a systematic study of linear models for time series forecasting, with a focus on the role of characteristic roots in temporal dynamics. We begin by analyzing the noise-free setting, where we show that characteristic roots govern long-term behavior and explain how design choices such as instance normalization and channel independence affect model capabilities. We then extend our analysis to the noisy regime, revealing that models tend to produce spurious roots. This leads to the identification of a key data-scaling property: mitigating the influence of noise requires disproportionately large training data, highlighting the need for structural regularization. To address these challenges, we propose two complementary strategies for robust root restructuring. The first uses rank reduction techniques, including Reduced-Rank Regression and Direct Weight Rank Reduction, to recover the low-dimensional latent dynamics. The second, a novel adaptive method called Root Purge, encourages the model to learn a noise-suppressing null space during training. Extensive experiments on standard benchmarks demonstrate the effectiveness of both approaches, validating our theoretical insights and achieving state-of-the-art results in several settings. Our findings underscore the potential of integrating classical theories for linear systems with modern learning techniques to build robust, interpretable, and data-efficient forecasting models.

Pascal Janetzky, Tobias Schlagenhauf, Stefan Feuerriegel

A common challenge in continual learning (CL) is catastrophic forgetting, where the performance on old tasks drops after new, additional tasks are learned. In this paper, we propose a novel framework called ReCL to slow down forgetting in CL. Our framework exploits an implicit bias of gradient-based neural networks due to which these converge to margin maximization points. Such convergence points allow us to reconstruct old data from previous tasks, which we then combine with the current training data. Our framework is flexible and can be applied on top of existing, state-of-the-art CL methods. We further demonstrate the performance gain from our framework across a large series of experiments, including two challenging CL scenarios (class incremental and domain incremental learning), different datasets (MNIST, CIFAR10, TinyImagenet), and different network architectures. Across all experiments, we find large performance gains through ReCL. To the best of our knowledge, our framework is the first to address catastrophic forgetting by leveraging models in CL as their own memory buffers.

Tobias Schlagenhauf, Niklas Burghardt

This paper addresses the ability to enable machines to automatically detect failures on machine tool components as well as estimating the severity of the failures, which is a critical step towards autonomous production machines. Extracting information about the severity of failures has been a substantial part of classical, as well as Machine Learning based machine vision systems. Efforts have been undertaken to automatically predict the severity of failures on machine tool components for predictive maintenance purposes. Though, most approaches only partly cover a completely automatic system from detecting failures to the prognosis of their future severity. To the best of the authors knowledge, this is the first time a vision-based system for defect detection and prognosis of failures on metallic surfaces in general and on Ball Screw Drives in specific has been proposed. The authors show that they can do both, detect and prognose the evolution of a failure on the surface of a Ball Screw Drive.

Tobias Schlagenhauf, Magnus Landwehr, Juergen Fleischer

Using machine learning (ML) techniques in general and deep learning techniques in specific needs a certain amount of data often not available in large quantities in technical domains. The manual inspection of machine tool components and the manual end-of-line check of products are labor-intensive tasks in industrial applications that companies often want to automate. To automate classification processes and develop reliable and robust machine learning-based classification and wear prognostics models, one needs real-world datasets to train and test the models. The dataset is available under https://doi.org/10.5445/IR/1000129520.

Tobias Schlagenhauf, Faruk Yildirim, Benedikt Brückner

Training deep learning models in technical domains is often accompanied by the challenge that although the task is clear, insufficient data for training is available. In this work, we propose a novel approach based on the combination of Siamese networks and radial basis function networks to perform data-efficient classification without pretraining by measuring the distance between images in semantic space in a data-efficient manner. We develop the models using three technical datasets, the NEU dataset, the BSD dataset, and the TEX dataset. In addition to the technical domain, we show the general applicability to classical datasets (cifar10 and MNIST) as well. The approach is tested against state-of-the-art models (Resnet50 and Resnet101) by stepwise reduction of the number of samples available for training. The authors show that the proposed approach outperforms the state-of-the-art models in the low data regime.

Tobias Schlagenhauf, Tim Brander, Juergen Fleischer

This paper provides a novel approach to stitching surface images of rotationally symmetric parts. It presents a process pipeline that uses a feature-based stitching approach to create a distortion-free and true-to-life image from a video file. The developed process thus enables, for example, condition monitoring without having to view many individual images. For validation purposes, this will be demonstrated in the paper using the concrete example of a worn ball screw drive spindle. The developed algorithm aims at reproducing the functional principle of a line scan camera system, whereby the physical measuring systems are replaced by a feature-based approach. For evaluation of the stitching algorithms, metrics are used, some of which have only been developed in this work or have been supplemented by test procedures already in use. The applicability of the developed algorithm is not only limited to machine tool spindles. Instead, the developed method allows a general approach to the surface inspection of various rotationally symmetric components and can therefore be used in a variety of industrial applications. Deep-learning-based detection Algorithms can easily be implemented to generate a complete pipeline for failure detection and condition monitoring on rotationally symmetric parts.

Tobias Schlagenhauf, Chenwei Sun, Jürgen Fleischer

The lack of reliable large datasets is one of the biggest difficulties of using modern machine learning methods in the field of failure detection in the manufacturing industry. In order to develop the function of failure classification for ball screw surface, sufficient image data of surface failures is necessary. When training a neural network model based on a small dataset, the trained model may lack the generalization ability and may perform poorly in practice. The main goal of this paper is to generate synthetic images based on the generative adversarial network (GAN) to enlarge the image dataset of ball screw surface failures. Pitting failure and rust failure are two possible failure types on ball screw surface chosen in this paper to represent the surface failure classes. The quality and diversity of generated images are evaluated afterwards using qualitative methods including expert observation, t-SNE visualization and the quantitative method of FID score. To verify whether the GAN based generated images can increase failure classification performance, the real image dataset was augmented and replaced by GAN based generated images to do the classification task. The authors successfully created GAN based images of ball screw surface failures which showed positive effect on classification test performance.

Tobias Schlagenhauf, Yefeng Xia, Jürgen Fleischer

Working with images, one often faces problems with incomplete or unclear information. Image inpainting can be used to restore missing image regions but focuses, however, on low-level image features such as pixel intensity, pixel gradient orientation, and color. This paper aims to recover semantic image features (objects and positions) in images. Based on published gated PixelCNNs, we demonstrate a new approach referred to as quadro-directional PixelCNN to recover missing objects and return probable positions for objects based on the context. We call this approach context-based image segment labeling (CBISL). The results suggest that our four-directional model outperforms one-directional models (gated PixelCNN) and returns a human-comparable performance.