Learning to Localize Objects with Structured Output Regression

By Matthew B. Blaschko, Christoph H. Lampert. 2008

380 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

You Only Look Once: Unified, Real-Time Object Detection (2015, methodology): “Then, classifiers [35, 21, 13, 10] or localizers [1, 31] are used to identify objects in the feature space.”
Structured class-labels in random forests for semantic image labelling (2011, methodology): “The work in [4] introduces a generalization of a support vector machine (SVM) for structured output regression, used to predict bounding boxes for the task of object localization.”
Max-Margin Object Detection (2015, result): “The best previous results on this dataset achieve an accuracy of 98.2% at equal error rate [2].”
Modeling Occlusion by Discriminative AND-OR Structures (2013, background): “For the task of view estimation, to cope with the problem that objects are correctly localized but by wrong view branches, we extend the loss function L in [1] as:”
Robobarista: Object Part Based Transfer of Manipulation Trajectories from Crowd-Sourcing in 3D Pointclouds (2015, background): “Thus, rather than relying on scene understanding techniques [7, 31, 15], we”
Ieee Transactions on Pattern Analysis an Machine Intelligence Structured Learning of Human Interactions in Tv Shows (positive): “Blaschko and Lampert [20] used it to learn a mapping between images and object bounding boxes, extending their approach in [21] to model local and global context information.”
Struck: Structured output tracking with kernels (2011, positive): “For example, as in [4] we choose to base this loss function on bounding box overlap, and use”
A coarse-to-fine approach for fast deformable object detection (2011, background): “This procedure, which is equivalent to constraint generation [35] or mining of hard negatives [3], is guarantee to end in polynomial time provided that the set of support vectors (i.e. the examples violating the margin at the optimum) can fit in memory.”
The Lovasz-Softmax Loss: A Tractable Surrogate for the Optimization of the Intersection-Over-Union Measure in Neural Networks (2017, methodology): “[3] optimize the Jaccard loss in a structured output SVM, but are only able to do so with a branch-and-bound optimization over bounding boxes and not full segmentations.”
Cascaded pose regression (2010, background): “However, as with standard classification efficient search/inference techniques are still required [20, 4].”
Class segmentation and object localization with superpixel neighborhoods (2009, background): “[3] have shown that it is feasible to search all possible sub-windows of an image for an object using branch and bound and a structured classifier whose output is a bounding box.”
Multiple Choice Learning: Learning to Produce Multiple Structured Outputs (2012, background): “For instance, [4] trains an SSVM for object detection implicitly defining a taskadapted loss, L(Y, ŷ) = miny∈Y `(y, ŷ).”
Loss Max-Pooling for Semantic Image Segmentation (2017, background): “In separate though related research fields, focus was put on directly optimizing the target measures like Area under curve (AUC), Intersection over Union (IoU or Jaccard Index) or Average class (AC) [1, 4, 36, 37].”
Latent hierarchical structural learning for object detection (2010, background): “An object is represented by a mixture of hierarchical tree models where the nodes represent object parts.”
Feature Mining for Localised Crowd Counting (2012, methodology): “An alternative to the multi-output ridge regression model is the structural Support Vector Machine [13], which has been applied to pose estimation [12] and object detection [3].”
Structured Forests for Fast Edge Detection (2013, background): “Structured learning: Structured learning addresses the problem of learning a mapping where the input or output space may be arbitrarily complex representing strings, sequences , graphs, object pose, bounding boxes etc. [35] [34] [2].”