Random Forests for Real Time 3D Face Analysis

By Gabriele Fanelli, Matthias Dantone, Juergen Gall, Andrea Fossati, Luc Van Gool. 2012

594 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

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

RGB-D datasets using microsoft kinect or similar sensors: a survey (2016, methodology): “9 From left to right: RGB images, depth images and depth images with annotations from Biwi Kinect Head Pose Dataset [25]”
OctNet: Learning Deep 3D Representations at High Resolutions (2016, methodology): “We evaluated our OctNet also on the Biwi Kinect Head Pose Database [11] as an additional experiment on 3D pose estimation.”
Learning an appearance-based gaze estimator from one million synthesised images (2016, methodology): “It has since been used in a number of systems, including learning-by-synthesis for head pose estimation [Fanelli et al. 2013], real-time facial performance capture [Li et al. 2013], and learning a person-specific face model for gaze estimation [Mora and Odobez 2012].”
Learning-by-Synthesis for Appearance-Based 3D Gaze Estimation (2014, background): “It enables to produce a large amount of training data as done in recent works on human body and head pose estimation [28, 8].”
Energy-Based Models for Deep Probabilistic Regression (2020, methodology): “We utilize the BIWI [12] dataset, specifically the processed dataset provided by Yang et al. [59], in which the images have been cropped to faces detected using MTCNN [61].”
SUN RGB-D: A RGB-D scene understanding benchmark suite (2015, unknown): “There are many existing RGB-D datasets available [54] [47] [1] [25] [44] [60] [49] [45] [29] [66] [57] [52] [46] [16] [21] [73] [35] [67] [3] [41] [10] [63] [42] [64] [65] [48] [12] [33] [8] [50] [26] [6].”
Head pose estimation via probabilistic high-dimensional regression (2015, methodology): “While the above mentioned methods mainly use color information, the release of depth sensors has led to the development of methods that rely on depth [8] or on RGB-D data [23].”
A Survey on Human Motion Analysis from Depth Data (2013, methodology): “Parts of this section appeared in [42].”
Body Parts Dependent Joint Regressors for Human Pose Estimation in Still Images (2014, background): “…reveals that our approach achieves a body part detection accuracy on a par with the state-of-the-art on the sports dataset and greatly outperforms [5] in terms of joint localization accuracy on the FashionPose dataset, while achieving a run time of a couple of images per second on the LSP dataset.”
Head pose estimation in the wild using Convolutional Neural Networks and adaptive gradient methods (2017, unknown): “with low-resolution images from monocular cameras [18], obtaining the best reported result on the Biwi Kinect Head Pose dataset [19].”
A Comprehensive Analysis of Deep Regression (2018, unknown): “The Biwi head-pose data set [43] consists of over 15 , 000 RGB-D images corresponding to video recordings of 20 people (16 men and 4 women, some of whom recorded twice) using a Kinect camera.”
Gradient boosting machines, a tutorial (2013, background): “…most prominent examples of such machine-learning ensemble techniques are random forests (Breiman, 2001) and neural network ensembles (Hansen and Salamon, 1990), which have found many successful applications in different domains (Liu et al., 2004; Shu and Burn, 2004; Fanelli et al., 2012; Qi, 2012).”
3D Head Pose Estimation for TV Setups (2013, methodology): “The 3D pose estimation algorithm is based on the approach in [7][10] and implemented in the PCL library [2].”
FSA-Net: Learning Fine-Grained Structure Aggregation for Head Pose Estimation From a Single Image (2019, methodology): “Three popular datasets for head pose estimation were adopted in the experiments: the 300W-LP [48], AFLW2000 [48], and BIWI [13] datasets, The 300W-LP dataset [48] was derived from the 300W dataset [37] which unifies several datasets for face alignment with 68 landmarks.”