Multi-PIE

By Ralph Gross, Iain A. Matthews, J. Cohn, Takeo Kanade, Simon Baker. 2010

2,100 citations recorded by the source corpus.

A close relationship exists between the advancement of face recognition algorithms and the availability of face databases varying factors that affect facial appearance in a controlled manner. The CMU PIE database has been very influential in advancing research in face recognition across pose and illumination. Despite its success the PIE database has several shortcomings: a limited number of subjects, single recording session and only few expressions captured. To address these issues we collected the CMU Multi-PIE database. It contains 337 subjects, imaged under 15 view points and 19 illumination conditions in up to four recording sessions. In this paper we introduce the database and describe the recording procedure. We furthermore present results from baseline experiments using PCA and LDA classifiers to highlight similarities and differences between PIE and Multi-PIE.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

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

In Signal Processing

Citations in the corpus

Using Fast Weights to Attend to the Recent Past (2016, methodology): “To further investigate the beneﬁts of using fast weights in the multi-level visual attention model, we performed facial expression recognition tasks on the CMU Multi-PIE face database [Gross et al., 2010].”
Weakly-supervised Disentangling with Recurrent Transformations for 3D View Synthesis (2015, methodology): “The encoder network for the Multi-PIE dataset used two convolution-relu layers with stride 2 and 2-pixel padding, followed by one fully-connected layer: 5 ˆ 5 ˆ 64 ´ 5 ˆ 5 ˆ 128 ´ 1024 .”
Discovering Hidden Factors of Variation in Deep Networks (2014, neutral): “The Multi-PIE datasets [24] consists of 754,200 high-resolution color images of 337 subjects.”
Learning Generative Models with Visual Attention (2013, methodology): “After ConvNet based approximate inference gives a good initialization, starting from step 5, 5 iterations of 20 leap-frog steps of HMC are used to sample the posterior and slightly improves the results.”
The MegaFace Benchmark: 1 Million Faces for Recognition at Scale (2015, background): “While, some companies have access to massive photo collections, e.g., Google in [24] trained on 200 Million pho- tos of 8 Million people (and more recently on 500M of 10M), these datasets are not available to the public and were used only for training and not testing.”
Beyond Face Rotation: Global and Local Perception GAN for Photorealistic and Identity Preserving Frontal View Synthesis (2017, methodology): “Our method is evaluated on MultiPIE [10], a large dataset with 750 , 000+ images for face recognition under pose, illumination and expression changes.”
Naive-Deep Face Recognition: Touching the Limit of LFW Benchmark or Not? (2015, background): “The LFW benchmark [8] is intended to test the recognition system’s performance in unconstrained environment, which is considerably harder than many other constrained dataset (e.g., YaleB [6] and MultiPIE [7]).”
Surpassing Human-Level Face Verification Performance on LFW with GaussianFace (2014, methodology): “The source-domain datasets include four different types of datasets as follows: Multi-PIE [19].”
A Semi-automatic Methodology for Facial Landmark Annotation (2013, background): “The remainder of this paper is structured as follows: Section 2 provides a brief overview of the existing facial databases.”
RASL: Robust alignment by sparse and low-rank decomposition for linearly correlated images (2010, background): “Recently, [38] proposed an image denoising algorithm based on low-rank matrix completion.”
Disentangled and Controllable Face Image Generation via 3D Imitative-Contrastive Learning (2020, methodology): “In addition, we follow the previous methods to evaluate our method on the Mul-tiPIE [15] images.”
Extensive Facial Landmark Localization with Coarse-to-Fine Convolutional Network Cascade (2013, methodology): “The images and annotations come from AFW, LFPW, HELEN, and IBUG [6, 1, 12, 7, 8].”
Controllable Continuous Gaze Redirection (2020, background): “Some publicly available gaze datasets, such as MPIIGaze [43], Gaze Capture [17], UT Multi-View [30] or CMU Multi-Pie [11], only provide low-resolution images and would therefore introduce an unwanted bias towards low-quality images.”
Accurate 3D Face Reconstruction With Weakly-Supervised Learning: From Single Image to Image Set (2019, methodology): “To train C-Net, we construct an image corpus using 300W-LP [57], Multi-PIE [17] and part of our in-house face recognition dataset.”
Multi-View Perceptron: a Deep Model for Learning Face Identity and View Representations (2014, methodology): “The number of neurons in MVP can be expressed as 32× 32− 512− 512(10)− 512(10)− 1024−32×32[7], where the input and output images have the size of 32×32, [7] denotes the length of the view label vector (v), and (10) represents that the third and forth layers have ten random neurons.”
LSDT: Latent Sparse Domain Transfer Learning for Visual Adaptation (2016, neutral): “The CMU Multi-PIE face dataset [23] is a comprehensive”
Geometry-Contrastive GAN for Facial Expression Transfer. (2018, methodology): “To evaluate the proposed GC-GAN, experiments have been conducted on four popular facial datasets: Multi-PIE [9], CK+ [16], BU-4DFE [28] and CelebA [15].”
Efficient $l_{1}$ -Norm-Based Low-Rank Matrix Approximations for Large-Scale Problems Using Alternating Rectified Gradient Method (2015, methodology): “In the experiments, we used 830 images having five different illuminations for 166 people from the Multi-PIE face database [27], which were resized to 100× 120 pixels.”