Bibliograph. Daten | Frisch, Jannik Daniel: Optimization of classification of car images using neural networks. Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Bachelorarbeit Nr. 25 (2017). 43 Seiten, englisch.
|
Kurzfassung | Neural networks are one of the most interesting topics of research in this time. The Fraunhofer IAO and the IMS both are interested in researching on this topic but also solving real life problems using neural networks. Industry leading companies such as Apple Inc. claim to use neural networks for various classification tasks in real life situations. Fraunhofer IAO would like to optimize processes for insurance companies by automatically classifying pictures of cars in their database. To learn more about the topic, this thesis deals with “how to use” neural networks on image classification tasks on car images. The work shows, that some classification tasks are easier to perform then others, and that only tuning network parameters is enough for some tasks to improve them, but not for others. The hardest task to solve was classifying images by make, model and model year of the car shown in the picture. To improve results, the complete application structure had to be changed to use a system of multiple neural networks, each trained for a single purpose, that the network could specialize on.
|
Volltext und andere Links | PDF (1784200 Bytes) Zugriff auf studentische Arbeiten aufgrund vorherrschender Datenschutzbestimmungen nur innerhalb der Fakultät möglich |
Abteilung(en) | Universität Stuttgart, Institut für Maschinelle Sprachverarbeitung
|
Betreuer | Vu, Jun.-Prof. Ngoc Thang; Kinz, MAximilian; Wohlfrom, Andreas |
Eingabedatum | 28. September 2018 |
---|