Auto-context and its application to high-level vision tasks

By Zhuowen Tu. 2008

291 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Relation Networks for Object Detection (2017, background): “Object relation in post-processing Most early works use object relations as a post-processing step [12, 17, 46, 47, 36, 17].”
Holistically-Nested Edge Detection (2015, background): “…response fusion [38, 30, 28], (3) engagement of different levels of visual perception [16, 25, 37, 15] such as mid-level Gestalt law information [7], (4) structural information [6] and context [36], (5) holistic image prediction [23], (6) 3D geometry [13], and (7) occlusion boundaries [14].”
Conditional Random Fields as Recurrent Neural Networks (2015, background): “Lack of such smoothness constraints can result in poor object delineation and small spurious regions in the segmentation output [31, 30, 14, 21].”
Human Pose Estimation with Iterative Error Feedback (2015, methodology): “The autocontext work of [41] is also related and iteratively computes label heatmaps by concatenating the image with the heatmaps previously predicted.”
Dense Recurrent Neural Networks for Scene Labeling (2018, methodology): “rations (layers). For example, Pinheiro et al.[17] attachs the RGB raw data with the output of the Convolutional Neural Network (CNN) to produce the input for the same CNN in the next layer. Tu et al.[28] augments the patch feature with the output of the classiﬁer to be the input for the next iterat”
Learning the easy things first: Self-paced visual category discovery (2011, background): “Various forms of context have been explored in the object recognition literature; several methods model the cooccurrence and relative spatial layout of objects to improve predictions (e.g., [20] [5] [15] [22] [9]).”
Beyond pixels: A comprehensive survey from bottom-up to semantic image segmentation and cosegmentation (2015, methodology): “Guo and Hoiem [119] chose to use Auto-Context [120] to incorporate the detector response in their system.”
VoxResNet: Deep Voxelwise Residual Networks for Volumetric Brain Segmentation (2016, unknown): “Furthermore, in order to harness the integration of high-level context information, implicit shape information and original low-level image appearance for improving recognition performance, we formulate the learning process as an auto-context algorithm [35].”
Exploring Person Context and Local Scene Context for Object Detection (2015, methodology): “Such contextual rescoring is an instantiation of the more general idea of autocontext [29], an iterative procedure where predictions from a previous iteration are used to provide contextual features for the next iteration.”
H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation From CT Volumes (2018, methodology): “Specifically, 3D DenseUNet is integrated with the 2D DenseUNet by the way of auto-context [26] mechanism, which is a general form of stacked generality [27].”
Few-Shot Generalization for Single-Image 3D Reconstruction via Priors (2019, methodology): “While the notion of iterative prediction for better accuracy has been proposed before [19, 15, 2], the connection to few-shot learning in this context is new.”
Decision Forests: A Unified Framework for Classification, Regression, Density Estimation, Manifold Learning and Semi-Supervised Learning (2012, background): “The basic idea stems from the work on Probabilistic Boosting Trees [104] and autocontext [105].”
DeepContext: Context-Encoding Neural Pathways for 3D Holistic Scene Understanding (2016, background): “The role of context has been studied extensively in computer vision [27, 41, 5, 4, 3, 7, 18, 37, 33, 35, 34, 19, 12, 9, 43, 1, 39, 24, 10, 42, 20, 21].”
A Tree-Based Context Model for Object Recognition (2012, methodology): “Tu [29] uses both image patches and their probability maps estimated from classifiers to learn a contextual model, and iteratively refines the classification results by propagating the contextual information.”
Top-Down Learning for Structured Labeling with Convolutional Pseudoprior (2015, methodology): “For example, cascade models [8,9,10] in the spirit of stacking [11], are proposed to take the outputs of classifiers of the current layer as additional features for the next classifiers in the cascade.”
Efficient Human Pose Estimation from Single Depth Images (2013, methodology): “Tu's " autocontext " was used in [27] to obtain a coarse body part labeling.”
Tracking Objects as Points (2020, methodology): “aps. PoseFix [29] generates heatmaps that simulate test errors for human pose reﬁnement. Our tracking-conditioned detection framework is inspired by these works. A rendered heatmap of prior keypoints [4,11,29,44] is especially appealing in tracking for two reasons. First, the information in the prev”
Layered object detection for multi-class segmentation (2010, methodology): “In contrast, multi-class segmentation models have largely been built on top of Markov Random Field (MRF) models which enforce smoothness across pixel labels [6, 21, 13, 20, 22].”