• Datum:  20.10.2023
  • Uhrzeit:  09:00 bis 16:00 Uhr
  • Ort: Online
  • Sprache:  Deutsch/English

Fachsymposium "Artificial Intelligence for Life"

Grafik zu KI und Lebenswissenschaften

Register here to the symposium | Hier können Sie sich für das Symposium registrieren

Ein spannendes und abwechslungsreiches Programm rund um das Thema der
Künstlichen Intelligenz.

Das englischsprachige Symposium ist kostenlos und auch für externe Teilnehmer (ohne KI-Vorwissen) zugänglich. Geplant ist ein ganzer Tag mit verschiedenen Themen, die im weitesten Sinne mit KI und Lebenswissenschaften zu tun haben.

Key Note: Dr. Martin Junghans, CTO Innovation Studios, IBM Deutschland GmbH
Scale and accelerate the impact of AI with foundation models
Präsentation von Herrn Dr. Junghans als pdf

Nora Gourmelon, FAU Erlangen-Nürnberg
AI-based Remote Sensing Applications
What do bird populations, glaciers and water consumption have in common? We can monitor them using remote sensing. To reduce manual effort, AI-based methods are being developed to take over this task

Siming Bayer, FAU Erlangen-Nürnberg
Untangling the unseen: Hybrid AI approach tackling inverse problems of water and energy supplies
highlights the merger of traditional methods and AI in solving leakage detection and localization in water and energy networks. Using a real-world example, the talk showcases the potential of hybrid AI models, emphasizing the importance of interdisciplinary collaboration in water and energy management.
Emanuel Jakobowsky, Technical University of Applied Sciences Ingolstadt
Detecting and counting wheat spike heads from UAV-based images using deep convolutional neural networks
Manual detection and counting of wheat head spikes is essential for maximizing wheat yields but this task is also time-consuming and subject to human error. This talk presents a plot-level analysis using a Mask R-CNN for automatic detection and counting of wheat spike heads on UAV-based images.
Präsentation von Herrn Jakobowsky als pdf
  • Dimo Dimov, University of Applied Sciences Karlsruhe
    Exploring the Potential of Multi-Temporal Crop Canopy Models and Vegetation Indices from Pleiades Imagery for Yield Estimation
    In this paper, we demonstrate the capabilities of Pleiades-1a imagery for very high resolution (VHR) crop yield estimation by utilizing the predictor variables from the horizontal-spectral information, through Normalized Difference Vegetation Indices (NDVI), and the vertical-volumetric crop characteristics, through the derivation of Crop Canopy Models (CCMs), from the stereo imaging capacity of the satellite.  We further explore the spatiotemporal relationship of the CCMs and the NDVI for five observation dates during the growing season for eight selected crop fields in Germany with harvester-measured ground truth crop yield. Moreover, we explore different CCM normalization methods, as well as linear and non-linear regression algorithms, for the crop yield estimation.
    Präsentation von Herrn Dimov als pdf
Raphael Oefelein, HSWT Standort Triesdorf
3D Crop Modeling with Neural Radiance Fields and Hyperspectral Imaging for Vegetation Index Computation
In today’s agricultural landscape, numerous complex challenges such as sustainable resource utilization,
environmental impact mitigation, and the need for more efficient production methodologies
have emerged. In this context, the creation of digital twins for crop plants presents new opportunities
for agricultural digitization. Realistic simulations aim to assist farmers in basing their
management practices on data-driven insights, optimizing resource utilization, while simultaneously
minimizing environmental impacts1. The development of a lifelike 3D object stands as a
pivotal step in the evolution of a comprehensive digital twin.
Andreas Gilson, Fraunhofer IIS
Digitizing Reality: AI as a Key Ingredient in the Creation of a complex Digital Twin
Digital Twins of real-world objects and processes continue to gain importance in various fields of application. A digital representation of a plant can be an invaluable asset for e.g. researchers and breeders and in the future also for producers. The ability to precisely capture the state of a plant at any time and analyze, assess and predict the phenotypic development - from architecture to fruit size – promises new insights. We describe the role of AI in the development of a digital twin of a cherry tree.
Dr. Sebastian Fritsch, HSWT Standort Triesdorf
More than ChatGPT: The benefits of Large Language Models for agriculture

This talk shows the current state of creating an AI co-pilot for agricultural operations, specifically crop protection. This is a topic where a lot of farmers struggle due to opaque legal guidelines and an abundance of information. The goal is to make legally relevant information available to users in a precise and streamlined way using large language models (LLMs). For these kinds of tasks it is however necessary to avoid the typical problems of LLMs: hallucinations, factually wrong answers and non-deterministic answers. One way to circumvent these problems is the application of retrieval augmented generation (RAG), in which a chatbot has access to knowledge stored in a database. In order to explore this topic we will first take a look at Langchain (www.langchain.com), one of the most important toolkits when it comes to LLMs. Langchain offers a wide variety of tools for data preparation like document loading and text splitting. After this step, text embedding models from OpenAI and HuggingFace are used to vectorize contextual information (agricultural guidelines) and store them in a vector database. The information stored in these databases can then be retrieved via Langchain and prompt engineering. This way the created chatbot (in theory) only gives factually correct answers to the user. After taking a closer look at RAG, a tool called Flowise will be presented, which is a visual programming tool that can be used to rapidly prototype chatbots using Langchain without the need for having extensive programming skills. Finally, we look at pitfalls and problems when creating a chatbot using RAG and how you can validate the answers it gives.
Präsentation von Herrn Dr. Fritsch als pdf

Prof. Dr. Dominik Grimm, HSWT Standort Straubing
Advancing Protein Engineering with Large Language Models
Accurate prediction of protein properties is an essential task in many areas of biotechnology, including enzyme engineering and protein-hybrid optoelectronics. In this talk, we will show the benefits of large language models for predicting protein thermophilicity and thermostability, and give an outlook on how these models will revolutionize the design of synthetic proteins.
Präsentation von Prof. Dr. Grimm als pdf



Weitere Veranstaltungen