Division of Microelectronic Systems Design (EMS)

Under way

Deep Learning-Based Scene Text Recognition in Mobile Outdoor Application

Type of work:

Master Thesis / Diplomarbeit

Assignment:

This Master’s thesis is conducted in cooperation with KAMAG Transporttechnik GmbH & Co.KG. The goal of the current thesis is the implementation of a scene text recognition in mobile outdoor application based on deep learning algorithm. It is expected to have a running prototype on Nvidia Jetson T2. The deep learning algorithm has to be implemented using Tensorflow or Pytorch. The other goal is to compare the existing non-machine learning based implementation with the deep learning based implementation from accuracy and runtime point of view.

Skills:

  • Interest in Deep Learning
  • Experience with Python or C/C++
  • Experience with PyTorch or Tensorflow

Background:

Scene text recognition is among the most important and challenging tasks in image-based sequence recognition. Scene text is text that appears in an image captured by a camera in an outdoor environment. The task is challenging as the text in scene images varies in shape, font, colour and position. The recognition of scene text is further complicated sometimes by non-uniform illumination and focus. Deep learning has been shown to be one of the most efficient approaches to provide a solution for the scene text recognition problem.

Supervisor:

V. Rybalkin, M. Tekleyohannes

Student:

Ebin Zacharias

Year:

2020

Zum Seitenanfang