Real Time Head Pose Estimation from Consumer Depth Cameras

By Gabriele Fanelli, Thibaut Weise, Juergen Gall, Luc Van Gool. DAGM-Symposium, 2011

365 citations recorded by the source corpus.

We present a system for estimating location and orientation of a person's head, from depth data acquired by a low quality device. Our approach is based on discriminative random regression forests: ensembles of random trees trained by splitting each node so as to simultaneously reduce the entropy of the class labels distribution and the variance of the head position and orientation. We evaluate three different approaches to jointly take classification and regression performance into account during training. For evaluation, we acquired a new dataset and propose a method for its automatic annotation.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/0f31704c65b57994ceb8e0b0f4733b113dd2ce86

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

OpenFace 2.0: Facial Behavior Analysis Toolkit (2018, neutral): “forests [25] 7.”
Alternating Regression Forests for Object Detection and Pose Estimation (2013, methodology): “As in [8], we split the data into 18 training sequences and 2 testing sequences.”
RT-GENE: Real-Time Eye Gaze Estimation in Natural Environments (2018, background): “CMU Multi-Pie [18] RGB Camera frame 68 Facial landmarks 755,370 ≈300cm X - All BIWI [13] RGB-D Camera frame Head pose vector ≈15,500 100cm X - All ICT 3D Head pose [2] RGB-D Camera frame Head pose vector 14,000 ≈100cm X - All Deep Head Pose [38] RGB-D Camera frame Head pose vector 68,000 ≈200-800cm ”
3D Face Pose Tracking using Low Quality Depth Cameras (2013, methodology): “We rather choose to perform the evaluation of our method mainly with the Biwi Kinect Head Pose Database (Fanelli et al., 2011) which contains 24 sequences of 24 different persons.”
3D Head Pose Estimation for TV Setups (2013, methodology): “The training stage is based on the BIWI dataset [10] containing over 15000 images of 20 people (6 females and 14 males).”
RGBD Datasets: Past, Present and Future (2016, background): “Biwi Kinect Head Pose Database [28] 20 Kinect v1 People moving their heads in different directions 3D position of the head and its rotation ’11”
Robots for humanity: using assistive robotics to empower people with disabilities (2013, methodology): “Using the system in [5] with the Kinect sensor, we track the 3D position and orientation of Henry’s head in real-time, and use this as a pointer for objects in the world.”
Random Forests for Real Time 3D Face Analysis (2012, methodology): “Inspired by the works (Gall and Lempitsky 2009; Criminisi et al. 2010), we have shown that regression forests can be used for real time head pose estimation from depth data (Fanelli et al. 2011a, 2011b).”
FSA-Net: Learning Fine-Grained Structure Aggregation for Head Pose Estimation From a Single Image (2019, methodology): “On the other hand, it is also very popular to adopt different modalities to compensate for the loss of information [14, 27, 9, 29].”
OpenFace: An open source facial behavior analysis toolkit (2016, neutral): “forests [22] 9.”
Computer Vision and Image Understanding (2016, methodology): “RF represents the method described in Fanelli et al. (2011) Fig.”
3D Constrained Local Model for rigid and non-rigid facial tracking (2012, methodology): “We used the Biwi Kinect head pose dataset [14].”
Head pose estimation in the wild using Convolutional Neural Networks and adaptive gradient methods (2017, neutral): “As it is possible to observe from Figures 5, 6 and 12, Adam and RMSProp converge to a very low loss value during the ﬁrst epochs.”
Real-time facial feature detection using conditional regression forests (2012, methodology): “To obtain the head pose, we train a regression forest similar to [13, 19].”
A Survey on Human Motion Analysis from Depth Data (2013, methodology): “Database for benchmarking head pose estimation from depth data [67].”