The virtually error-free segmentation and tracking of densely packed cells and cell nuclei is still a challenging task. Especially in low-resolution and low signal-to-noise-ratio microscopy images erroneously merged and missing cells are common segmentation errors making the subsequent cell tracking even more difficult. In 2020, we successfully participated as team KIT-Sch-GE (1) in the 5th edition of the ISBI Cell Tracking Challenge. With our deep learning-based distance map regression segmentation and our graph-based cell tracking, we achieved multiple top 3 rankings on the diverse data sets. In this manuscript, we show how our approach can be further improved by using another optimizer and by fine-tuning training data augmentation parameters, learning rate schedules, and the training data representation. The fine-tuned segmentation in combination with an improved tracking enabled to further improve our performance in the 6th edition of the Cell Tracking Challenge 2021 as team KIT-Sch-GE (2).
Programmatic Formation explores design as a responsive process.The study we present engages the complexity of the surroundings using parametric and generative design methods. It illustrates that responsiveness of designs can be achieved beyond geometric explorations.The parametric models can combine and respond simultaneously to design and its programmatic factors, such as performance-sensitive design-decisions, and constraints.We demonstrate this through a series of case studies for a housing tower.The studies explore the extent to which non-spatial parameters can be incorporated into spatial parametric dependencies in design.The results apply digital design and modeling, common to the curriculum of architecture schools, to the practical realm of building design and city planning.While practitioners are often slow to include contemporary design and planning methods into their daily work, the research illustrates how the incorporation of skills and knowledge acquired as part of university education can be effectively incorporated into everyday design and planning.
Supervised deep learning approaches for automated diagnosis support require datasets annotated by experts. Intra-annotator variability of a single annotator and inter-annotator variability between annotators can affect the quality of the diagnosis support. As medical experts will always differ in annotation details, quantitative studies concerning the annotation quality are of particular interest. A consistent and noise-free annotation of large-scale datasets by, for example, dermatologists or pathologists is a current challenge. In this paper, we categorize annotation noise in image segmentation tasks, present methods to simulate annotation noise, and examine the impact on the segmentation quality. Two novel automated methods to identify intra-annotator and inter-annotator inconsistencies based on uncertainty-aware deep neural networks are proposed. We demonstrate the benefits of our inspection methods such as focused reinspection of noisy annotations or the detection of generally different annotation styles using the biomedical ISIC 2017 Melanoma image segmentation dataset.
Summary: Reliable deep learning segmentation for microfluidic live-cell imaging requires comprehensive ground truth data. ObiWan-Microbi is a microservice platform combining the strength of state-of-the-art technologies into a unique integrated workflow for data management and efficient ground truth generation for instance segmentation, empowering collaborative semi-automated image annotation in the cloud. Availability and Implementation: ObiWan-Microbi is open-source and available under the MIT license at https://github.com/hip-satomi/ObiWan-Microbi, along documentation and usage examples. Contact: k.noeh@fz-juelich.de Supplementary information: Supplementary data are available online.
With the development of the smart grid, the number of recorded energy and power times series increases noticeably. This increase allows for the automation of smart grid applications such as load forecasting and load management. This automation, however, requires data that only represents the typical behavior of the system. To ensure that such data is available, detecting the anomalies often present in recorded data is important. As a result, anomaly detection methods are a recent research topic. However, their development is often limited by undefined anomaly characteristics and a lack of labeled anomalous data. To overcome this challenge, we propose a method that generates synthetic anomalies based on real-world anomalies that can be inserted into energy and power time series. For this, we analyze real energy and power time series to identify four types of commonly occurring anomalies. Given the identified anomaly types, we formally model each type and use these models to insert synthetic anomalies of each type into arbitrary energy or power time series. We show that our method is not only capable of generating synthetic anomalies with real-world properties, but also beneficial for training supervised anomaly detection methods. CCS CONCEPTS• Computing methodologies → Anomaly detection.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.