Logo des Repositoriums

Auflistung nach Autor:in "von Szadkowski, Kai"

1 - 3 von 3
  • Konferenzbeitrag
    Evaluating synthetic vs. real data generation for AI-based selective weeding
    (43. GIL-Jahrestagung, Resiliente Agri-Food-Systeme, 2023) Iqbal, Naeem; Bracke, Justus; Elmiger, Anton; Hameed, Hunaid; von Szadkowski, Kai
    Synthetic data has the potential to reduce the cost for ML training in agriculture but poses its own set of problems compared to real data acquisition. In this work, we present two methods of training data acquisition for the application of machine vision algorithms in the use case of selective weeding. Results from ML experiments suggest that current methods for generating synthetic data in the field of agriculture cannot fully replace real data but may greatly reduce the quantity of real data required for model training.
  • Konferenzbeitrag
    Instance-level augmentation for synthetic agricultural data using depth maps
    (43. GIL-Jahrestagung, Resiliente Agri-Food-Systeme, 2023) Wübben, Henning; Butz, Raphaela; von Szadkowski, Kai; Barenkamp, Marco
    Image augmentation is a key component in computer vision pipelines. Its techniques utilize different levels of data annotation. A lack of methods can be observed when it comes to data that supplies depth maps, in particular synthetic data. We propose a novel augmentation method named DepthAug that utilizes depth annotations in image data and examine its performance in the context of object detection tasks. Results show a boost in MAP score performance compared to previous related methods.
  • Konferenzbeitrag
    Towards selective hoeing depending on evaporation from the soil
    (43. GIL-Jahrestagung, Resiliente Agri-Food-Systeme, 2023) Manss, Christoph; von Szadkowski, Kai; Bald, Janis; Richard, David; Scholz, Christian; König, Daniel; Ruckelshausen, Arno
    This paper presents how to generate an artificial dataset to test different hoeing rules. Therefore, images that have been obtained on two days of a field trial are analysed to infer weed and crop sizes. Then, weather data from 2021 and 2022 is gathered from open-source data for 100 synthetically generated fields. The generated dataset is then used to test hoeing rules that are conditioned to keep as much moisture in the soil as possible. The analysis with these hoeing rules indicates that much less hoeing would be applied if the proposed hoeing rules are used.