Structured class-labels in random forests for semantic image labelling

By Peter Kontschieder, Samuel Rota Bulò, Horst Bischof, Marcello Pelillo. 2011

230 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

Fast Edge Detection Using Structured Forests (2014, methodology): “E.g., for stumps where θ = (k, τ) and h(x, θ) = [x(k) τ ], [20] advocates sampling √ d features where X = Rd and a single threshold τ per feature.”
Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding (2015, neutral): “4 n/a Structured Random Forests [20] n/a 51.”
Importance-Aware Semantic Segmentation for Autonomous Driving System (2017, methodology): “…methods focus on designing powerful handcrafted features and using Random Forest [Shotton et al., 2008; Brostow et al., 2008; Silberman et al., 2012] or boostingbased [Sturgess et al., 2009; Kontschieder et al., 2011; Zhang et al., 2010] classifiers for predicting the class of image pixels.”
Affordance detection of tool parts from geometric features (2015, methodology): “For supervised methods, stateof-the-art performance using structured random forests [21] applied over RGB-D data for simultaneous object segmentation and recognition has been reported in [13].”
Integrating parametric and non-parametric models for scene labeling (2015, background): “[11, 12, 18] modeled the higher order relations by considering patch/superpixel as a clique.”
Exemplar-Based Face Parsing (2013, background): “There are several recent scene parsing approaches in the literature that do not target faces, but nonetheless do share some aspects with our approach ([7, 8, 12, 19] to name just a few).”
Projection onto the Manifold of Elongated Structures for Accurate Extraction (2015, background): “It is inspired by the work of [20, 21] where the structured random forest framework is introduced for image labeling purposes, predicting for every pixel an image patch, instead of single pixel probabilities.”
SemiContour: A Semi-supervised Learning Approach for Contour Detection (2016, background): “Existing solutions [17, 12] can only handle fully labeled data, and are not applicable in our case that contains both unlabeled and labeled data.”
Quaddirectional 2D-Recurrent Neural Networks For Image Labeling (2015, background): “fully connected [6] or larger-clique [7][8] graphical models.”
Structured Forests for Fast Edge Detection (2013, methodology): “Similar to the task of semantic image labeling [20], the labels within a small image patch are highly interdependent, providing a promising candidate problem for our structured forest approach.”
Food-101 - Mining Discriminative Components with Random Forests (2014, background): “and proved successful for many vision applications, including object recognition [3,29,33], object detection [11] and semantic segmentation [22,33].”
Decision Forests: A Unified Framework for Classification, Regression, Density Estimation, Manifold Learning and Semi-Supervised Learning (2012, background): “The work in [54] presents a different approach, where the authors train a classification forest to predict “label patches,” as opposed to individual labels for each pixels.”
Dense Recurrent Neural Networks for Scene Labeling (2018, background): “. [24] formulated the unary and pairwise features in a 2ndorder CRF graphical model. Zhang et al.[40] and Roy et al. [23] built a fully connected graph to enforce higher order labeling coherence. [13][14][20] modeled the higher order consistency by considering patch/superpixel as a clique. He et al.”
Automatic Road Crack Detection Using Random Structured Forests (2016, background): “For example, in [24], researchers apply structured learning to semantic image labeling where image labels are also interdependent.”
EfficientPS: Efficient Panoptic Segmentation (2020, background): “Kontschieder et al (2011) exploit the inherent topological distribution of object classes to improve the performance, whereas Krähenbühl and Koltun (2011) improve segmentation by pairing CRFs with Gaussian edge potentials.”