Augmenting CRFs with Boltzmann Machine Shape Priors for Image Labeling

By Andrew Kae, Kihyuk Sohn, Honglak Lee, Eric Learned-Miller. 2013

194 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

Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation (2019, methodology): “In addition, [66] was evaluated on segmenting face images into 3 classes [35], whereas our models are evaluated on large-scale segmentation datasets such as Cityscapes [13], PASCAL VOC 2012 [15], and ADE20K [89].”
Convolutional Neural Fabrics (2016, methodology): “Moreover, [10, 19, 31] also include CRF and/or RBM models to encode spatial shape priors.”
Not All Pixels Are Equal: Difficulty-Aware Semantic Segmentation via Deep Layer Cascade (2017, background): “While early efforts focused on structural models with handcrafted features [15, 16, 34, 38], recent studies employ deep convolutional neural network (CNN) to learning strong representation, which improves segmentation accuracy significantly [3, 22, 23, 25, 40].”
Learning Time-Efficient Deep Architectures with Budgeted Super Networks (2017, methodology): “We propose to focus on two tasks: Image classification using MNIST and CIFAR-10 [11, 10] and Image Segmentation using the Part Labels dataset [9].”
Unsupervised Object Segmentation by Redrawing (2019, methodology): “A subpart of the funneled version [24] has been segmented and manually annotated [27], providing 2927 groundtruth masks.”
Deep learning for visual understanding (2017, unknown): “[101] combined the CRF and the RBM to model both local and global structure in face segmentation, which has consistently reduced the error in face labeling.”
FSGAN: Subject Agnostic Face Swapping and Reenactment (2019, methodology): “We also used the LFW Parts Labels set [19] with ∼ 3 k images labeled for face and hair segmentations, removing the neck regions using facial landmarks.”
Deep learning for visual understanding: A review (2016, background): “Kae et al. [169] combined the CRF and the RBM to model both local and global structure in face segmentation, which has consistently reduced the error in face labeling.”
Learning High-level Prior with Convolutional Neural Networks for Semantic Segmentation (2015, methodology): “pixel Average Precision (SAP) as noted in [14].”
Scene parsing using inference Embedded Deep Networks (2016, methodology): “Different from the way of some previous work combining deep learning methods with CRFs [2,31], which regards the deep learning methods as feature generative model to extract high-level hierarchical features and treats the CRFs as a label discriminative model to give the final category of every pixel”
From Facial Parts Responses to Face Detection: A Deep Learning Approach (2015, methodology): “Note that we combine the evaluation on ‘Hair+Beard’ to suit the ground truth provided by [14] (see Sec.”
A generative restricted Boltzmann machine based method for high-dimensional motion data modeling (2015, methodology): “In [22,23], RBM and CRF are combined for face labeling problem in a single image or video sequences.”
Learning Time/Memory-Efficient Deep Architectures with Budgeted Super Networks (2017, methodology): “Image classification has been tested on CIFAR-10 and CIFAR-100 [16] while the image segmentation has been performed on the Part Label dataset [15].”
Learning Structured Output Representation using Deep Conditional Generative Models (2015, methodology): “Following region labeling protocols [35, 13], we evaluate the performance of face parts labeling on the subset of LFW database [35], which contains 1 , 046 images that are labeled into 4 semantic categories, such as hair, skin, clothes, and background.”
Deep Learning for Semantic Part Segmentation with High-Level Guidance (2015, methodology): “Despite small theoretical differences, the major practical difference between our method and the aforementioned ones is that we do not use any superpixels, pooled features, or boundary signals, as Kae et al. (2013) .”
Top-Down Learning for Structured Labeling with Convolutional Pseudoprior (2015, methodology): “[26] also combines RBM and CRF to incorporate shape prior for face segmentation.”
Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs (2017, methodology): “Some methods incorporate higher order potentials in CRFs (Kohli et al., 2009) or model global shape priors with Restricted Boltzmann Machines (Li et al., 2013; Kae et al., 2013; Yang et al., 2014; Eslami et al., 2014).”
Bottom-Up and Top-Down Reasoning with Hierarchical Rectified Gaussians (2015, methodology): “ on LFW-Parts, following the protocol introduced in [25]. We show qualitative results in Fig. 10. We report both overall accuracy and error reduction with respect to a CRF baseline (as recommended by [25]). Our bottom-up baseline (CLVM 1) outperforms all published work. Adding topdown feedback (CLVM”
Segmentation of Both Diseased and Healthy Skin From Clinical Photographs in a Primary Care Setting (2018, background): “: 1) Labeled Faces in the Wild (LFW) [4], 2) Labeled Video Sequences (LVS) [5], and 3) a hand gesture recognition (HGR) dataset [6], [7], [8].”