Studienprojekt: Study Project: Product identification from images and videos (Part II) - Details

Studienprojekt: Study Project: Product identification from images and videos (Part II) - Details

Nachrichtenbox schließen
Sie sind nicht in Stud.IP angemeldet.

Allgemeine Informationen

Veranstaltungsname Studienprojekt: Study Project: Product identification from images and videos (Part II)
Untertitel
Veranstaltungsnummer 8.3341
Semester SoSe 2018
Aktuelle Anzahl der Teilnehmenden 12
Heimat-Einrichtung LE Cognitive Science
Veranstaltungstyp Studienprojekt in der Kategorie Offizielle Lehrveranstaltungen
Erster Termin Montag, 09.04.2018 10:00 - 12:00, Ort: 50/E04
Art/Form
Voraussetzungen The ideal participant in this project has some background in machine learning and computer vision, and a basic knowledge of neural networks. Some experience with TensorFlow or a similar framework would be of additional benefit.
SWS 3
Sprache Englisch
ECTS-Punkte 12

Räume und Zeiten

50/E04
Montag: 10:00 - 12:00, wöchentlich (12x)

Kommentar/Beschreibung

This is a continuation of the study project from last semester. Only students that have participated in the part I can subscribe for part II

The goal of this study project is to develop a computer vision system for product classification that can scale up to a huge number of classes. The ultimate goal would be the realization of a "Product Shazam" App, allowing to look up product information based on a single snapshot.
In the project, we will focus on deep learning techniques, like convolutional and recurrent neural networks. We will analyze recent developments in the field, evaluate selected approaches and check their potential for the envisioned system. Based on this analysis, we will develop a prototype system and test it on a real data set.
The project will be conducted in cooperation with a partner from local industry, who can provide a large set of training data and an existing system for baseline evaluation, as well as project management tools. Trainings on the relevant topics can be provided on demand.

In the project, we will focus on deep learning techniques, like convolutional and recurrent neural networks. We will analyze recent developments in the field, evaluate selected approaches and check
their potential for the envisioned system. Based on this analysis, we will develop a prototype system and test it on a real data set.

The project will be conducted in cooperation with a partner from local industry, who can provide a large set of training data and an existing system for baseline evaluation, as well as project management tools. Trainings on the relevant topics can be provided on demand.