This Doctoral Thesis investigates the significant role of data handling in the
practical application of deep learning techniques for object detection in highresolution
images. The study examines the impact of attention mechanisms
and introduces novel data processing methodologies, namely Artificial Size
Slicing Aided Fine Tuning (ASSAFT) and Artificial Size Slicing Aided Hyper
Inference (ASSAHI). Despite the potential of attention mechanisms observed in
medical imaging, the practical application of similar principles in the custommade
Tomato360 dataset does not prove to be beneficial. On the other hand,
a substantial improvement in object detection performance in the Tomato360
dataset was achieved through the newly proposed ASSAFT and ASSAHI techniques.
The research underlines the challenges of deploying deep learning techniques
in real-world scenarios; concretely, the final proposed solution is utilized
and evaluated for estimating crop yields in tomato greenhouses.
| ISBN: | 978-80-7678-191-7 |
|---|
| EAN: | 9788076781917 |
|---|
| Počet stran |
44 stran |
| Datum vydání |
11. 10. 2023 |
| Pořadí vydání |
první |
| Jazyk |
anglický |
| Vazba |
E-kniha |
| Autor: |
Alžběta Turečková |
| Nakladatelství |
Univerzita Tomáše Bati ve Zlíně |
| Tématická skupina |
5 - Technické vědy |
| Neprodejná publikace. Publikaci je možné poptávat zde: Volně dostupné na http://hdl.handle.net/10563/52453 |
|---|
Svaz českých knihkupců a nakladatelů