Pylon Model for Semantic Segmentation

By Victor S. Lempitsky, Andrea Vedaldi, Andrew Zisserman. 2011

163 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

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (2016, background): “Even thoughworks of increased sophistication have been proposed to model the hierarchical dependency [26], [27], [28] and/or high-order dependencies”
Minding the Gaps for Block Frank-Wolfe Optimization of Structured SVMs (2016, methodology): “In addition to images and their pixel-level annotations, Kolesnikov et al. (2014) released15 oversegmentations (superpixels) of the images precomputed with the SLIC algorithm (Achanta et al., 2012) and the unary features of each superpixel computed similarly to the work of Lempitsky et al. (2011).”
Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs (2014, background): “Even though works of increased sophistication have been proposed to model the hierarchical dependency (He et al., 2004; Ladicky et al., 2009; Lempitsky et al., 2011) and/or high-order dependencies of segments (Delong et al., 2012; Gonfaus et al., 2010; Kohli et al., 2009; Chen et al., 2013; Wang et…”
Feedforward semantic segmentation with zoom-out features (2014, methodology): “It is well known that training on imbalanced data without taking precautions can lead to poor results [10, 30, 22].”
Recurrent Convolutional Neural Networks for Scene Parsing (2013, methodology): “These models are then trained to maximize the likelihood of correct classiﬁcation given the features [1–7].”
Learning High-level Prior with Convolutional Neural Networks for Semantic Segmentation (2015, methodology): “At early stage, discriminative feature is usually hand-engineered, such works include [25, 19, 38] while classifiers vary from linear model, support vector machine to random forest.”
Recurrent Convolutional Neural Networks for Scene Labeling (2014, methodology): “, 2011 [7] HISTOGRAM OF VISUAL SIFT, HISTOGRAM OF RGB, HISTOGRAM OF LOCATIONS, “CONTOUR SHAPE” DESCRIPTOR.”
CRF Learning with CNN Features for Image Segmentation (2015, unknown): “[26] 81.”
Perceptually Inspired Layout-Aware Losses for Image Segmentation (2014, background): “[16] balance costs for false positives and false negatives in order to compensate for differences in prior probabilities.”
Closed-Form Approximate CRF Training for Scalable Image Segmentation (2014, unknown): “[22] – 75.”
Semantic Image Segmentation with Contextual Hierarchical Models (2015, unknown): “Pylon Model [69] 81:9% 72:4% 60”
Multi-scale context for scene labeling via flexible segmentation graph (2016, methodology): “Unlike these methods, we employ the covering relationship to capture visual context, and investigate this kind of context in different scales, without requiring any complex inference procedures to yield cleanly delineated predictions, such as sampling technique [23] or graph cut algorithms [42,43].”
Deep hierarchical parsing for semantic segmentation (2015, unknown): “ these approaches are mainly in terms of the visual features, form of the N-ary potentials and the the CRF modeling. A joint-CRF on multiple levels of an image segmentation hierarchy is formulated in [11]. It achieves better results than a ﬂat-CRF owing to the utilization of higher order contextual ”
Instance Segmentation of Indoor Scenes Using a Coverage Loss (2014, methodology): “Like many previous works in semantic segmentation [8] [4] [9], we use a heuristic to limit the search space by first performing a hierarchical segmentation of the image.”
Learning a Dictionary of Shape Epitomes with Applications to Image Labeling (2013, methodology): “Our models perform somewhat worse than the long-range model of [9] (unlike the MSRC case), and the segmentation tree model of [21], which however employs different image features.”
Scene parsing with Multiscale Feature Learning, Purity Trees, and Optimal Covers (2012, methodology): “4% / - < 600s Lempitsky et al. (2011) 81.”
Figure-ground segmentation by transferring window masks (2012, result): “While we outperform [1, 16, 31], the results of [27] are very impressive on this dataset.”