Computer Vision and Machine Learning
Here you can find an overview of current research projects within this area. More details and related publications can be found on the respective project pages. Contact information for the team leader can be found below.
Current Research Areas
Understanding Deep Learning
Deep neural networks are tremendously successful in many applications, but end-to-end trained networks often result in hard to understand black-box classifiers or predictors. In this Project, we face this problem. We develop methods, to determine whether a feature is relevant to the decision of a deep neural network. Further, we develop methods to influence which features will be considered by a neural network during training. This can help to learn more robust, trustworthy and fair deep neural networks.
Time frame: 2017 – 2023The AMMOD Project
The goal of the AMMOD project is to monitor biological diversity in Germany using new technologies. To this end, prototypes for so-called AMMOD stations are to be set up in German forests. These are equipped with sensors for recording animal sounds and plant emissions, with animal cameras for birds, mammals and insects, and with automated insect and pollen sample collectors for monitoring by DNA barcoding. These recordings will be used to generate a solid data pool that will enable the analysis of change and possible trends. For instance, we want to use the detected species to determine their population and record the change over a long period of time. The visual information will be analyzed, in part, by the the researchers at the Computer Vision Group in Jena. The Moth Scanner and the Novelty Detection, Life-Long Learning and Classification sub-tasks of the visual AMMOD System (visAMMOD) will be tackled here.
Time frame: 2019 – 2023Fine-grained Recognition
In this research area, we are developing methods that are able to automatically distinguish between very similar object categories. The algorithms learn from given pictures and their annotations the locations as well as the characteristics of relevant features. Applications for this research include automatic biodiversity monitoring.
Time frame: 2014 – 2023Lifelong Learning
Lifelong learning denotes a concept for continuous learning from data streams that typically become available during the application of a recognition system, e.g., additional images captured during a visual monitoring study. Instead of only applying a fixed model pre-trained on an initial dataset, lifelong learning aims at leveraging the additional data collected during an application in order to improve the recognition model over time. It incorporates active learning to exploit human feedback via annotations of unlabeled samples that have an impact on model training. Furthermore, incremental learning is required to update model parameters effectively and efficiently. In unconstrained environments and open-world scenarios, novelty detection also needs to be considered.
Time frame: 2012 – 2022Active Learning
Although labeled data lies at the very core of most computer vision systems, obtaining labeled data that is useful and reliable is commonly a crucial problem. To reduce the amount of manual labeling, active learning techniques aim at explicitely picking samples that are actually worth being labeled with respect to the problem on hand. In this area of research, we are interested in modeling the “worthyness” of an unlabeled sample and to apply our algorithms to human-in-the-loop recognition systems.
Time frame: 2012 – 2022Incremental Learning
Incremental learning refers to updating the parameters of a recognition model with additional (labeled) training data instead of training a new model from scratch with the extended set of training samples. It is also called continuous learning and is highly relevant for applications in which a stream of incoming data (structured in data chunks called experiences) is available and should be exploited for further improving the trained models. It is an important aspect for lifelong learning.
Time frame: 2012 – 2022Novelty Detection
In many important learning tasks, training examples are often available for only one class. Learning in this scenario is difficult, since training examples from “outside” are not available at all. This problem is known as one-class classification (OCC), novelty detection, and outlier detection, to name just a few. Our work in this area addresses the use of methods for deriving a set of suitable OCC scores and to apply OCC in an incremental learning framework.
Time frame: 2012 – 2022Contact
| Paul Bodesheim |
| Dr.-Ing. |
| Team Leader |
| paul.bodesheim@uni-jena.de |
| Room: 1218 |
| Phone: (+49) 3641 9 46410 |