Thema
Deep Learning based Gait Recognition by Kinect
Typ Master, IDP, Forschungspraxis
Betreuer Dr.-Ing. Mohammadreza Babaee
Tel.: +49 (0)89 289-28543
E-Mail: babaee@mmk.ei.tum.de
Sachgebiet Computer Vision
Beschreibung In intelligent visual surveillance, people are usually observed with distant from multiple cameras. Generally, face recognition techniques are used to identify the individuals. However, the shortcomings of this approach are low-resolution face image, due to large distances, unexpected view angle, and occlusion. Therefore, Gait is considered to be the most suitable biometrics alternative in visual surveillance.
Today, deep learning plays a key role in many computer vision problems such as image super resolution, image classification, and object recognition. Instead of hand crafted features such as SIFT, HOG, SURF, etc., deep learning algorithms extract a useful representation of the content of images.
In this project, the goal is to apply deep learning for gait recognition. People are imaged by Kinect that delivers us a color image and a depth map (i.e., RGB-D image). A sample n image of an RGB-D image of several people is presented below.

The idea is to apply a novel or existing deep learning algorithm to learn a new image representation from both RGB and Depth images in order to increase the accuracy of gait recognition. It is recommended to use one of the existing libraries for deep learning like caffe or torch for implementing the algorithm.

Reference:
• Preis, Johannes, et al. "Gait recognition with Kinect." 1st International Workshop on Kinect in Pervasive Computing. 2012.
• Hofmann, Martin, et al. "The TUM gait from audio, image and depth (GAID) database: Multimodal recognition of subjects and traits." Journal of Visual Communication and Image Representation 25.1 (2014): 195-206.
• Choudhury, Sruti Das, and Tardi Tjahjadi. "Robust view-invariant multiscale gait recognition." Pattern Recognition 48.3 (2015): 798-811.
• Joshi, Arun, Mr Shashi Bhushan, and Ms Jaspreet Kaur. "Gait recognition of human using SVM and BPNN classifiers." Int. J. Comput. Sci. Mobile Comput 3.1 (2014): 281-290.
Voraussetzung Preliminary knowledge in Machine learning and good programming skill in Matlab and C++ is highly required. For further questions, please write me an email to set an appointment.
Bewerbung If you are interested in this topic, we welcome the applications via the email address above. Please set the email subject to “<Type of application> application for topic 'XYZ'”, ex. “Master’s thesis application for topic 'XYZ'”, while clearly specifying why are you interested in the topic in the text of the message. Also make sure to attach your most recent CV (if you have one) and grade report.