Master Vortrag Rieger Head Pose Estimation Using Deep Learning

Master Vortrag Rieger Head Pose Estimation Using Deep Learning

от Ute Schmid -
Number of replies: 0

Am MI, 9.5.2018,  12:00 (das ist pünktlich um 12, also sine tempore) wird Ines Rieger ihre Masterarbeit verteidigen, die sie in Zusammenarbeit mit der Fraunhofer IIS Gruppe Intelligente Systeme erstellt hat. Gäste willkommen.


Ines Rieger (MA CitH): Head Pose Estimation Using Deep Learning

Head poses are an important mean of non-verbal human communication and thus a
crucial element in human-computer interaction. While computational systems have
been trained with various methods for head pose estimation in the recent years, ap-
proaches based on convolutional neural networks (CNNs) for image processing have
so far proven to be one of the most promising ones. This master’s thesis starts of
improving head pose estimation by reimplementing a recent CNN approach based on
the shallow LeNet-5. As a new approach in head pose estimation, this thesis focuses
on residual networks (ResNets), a subgroup of CNNs specifically optimized for very
deep networks. To train and test the approaches, the Annotated Facial Landmarks in
the Wild (AFLW) dataset and the Annotated Faces in the Wild (AFW) benchmark
dataset were used. The performance of the reimplemented network and the imple-
mented ResNets of various architectures were evaluated on the AFLW dataset. The
performance is hereby measured in mean absolute error and accuracy. Furthermore,
the ResNets with a depth of 18 layers were tested on the AFW dataset. The best
performance of all implemented ResNets was achieved by the 18 layer ResNet adapted
for an input size of 112 x 112 pixels. In comparison with the reimplemented network
and other state-of-the-art approaches, the best ResNet performs equal or better on the AFLW dataset and outperforms on the AFW dataset.