Recognition using regions

By Chunhui Gu, Joseph J. Lim, Pablo Arbeláez, Jitendra Malik. 2009

452 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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

In Media Technology

Citations in the corpus

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, methodology): “We start SGD at a learning rate of 0.001 (1/10th of the initial pre-training rate), which allows fine-tuning to make progress while not clobbering the initialization.”
Scalable Object Detection Using Deep Neural Networks (2013, background): “Some of these approaches build on bottom-up classless segmentation [9].”
What Makes for Effective Detection Proposals? (2015, background): “Sliding window classifiers scale linearly with the number of windows tested, and while single-scale detection requires classifying around 104 – 105 windows per image, the number of windows grows by an order of magnitude for multi-scale detection.”
Object-Contextual Representations for Semantic Segmentation (2019, methodology): “There exist many region-wise segmentation methods [1,2,21,20,64,50,2,60] that organize the pixels into a set of regions (usually super-pixels), and then classify each region to get the image segmentation result.”
Multiscale Combinatorial Grouping for Image Segmentation and Object Proposal Generation (2015, neutral): “Examples in this family are [4], [5].”
Full-Resolution Residual Networks for Semantic Segmentation in Street Scenes (2016, methodology): “Overview In Table 2 we compare our method to the best (published) performers on the Cityscapes leader board, namely LRR [20], Adelaide [23], and Dilation [55].”
How good are detection proposals, really? (2014, methodology): “gPbUCM [16] is a leading method for grouping pixels into objects.”
segDeepM: Exploiting segmentation and context in deep neural networks for object detection (2015, methodology): “[10] cast votes for the object’s location by using a Hough transform with a set of regions.”
Improving object detection with deep convolutional networks via Bayesian optimization and structured prediction (2015, methodology): “[18] proposed the R-CNN following the “recognition using regions” paradigm [19], which also inspired several previous state-of-the-art methods [44, 45].”
Hough Regions for Joining Instance Localization and Segmentation (2012, methodology): “[10] use regions as underlying reasoning for object detection, however they rely on a single over-segmentation of the image, which cannot recover from initial segmentation errors.”
Object detection using a max-margin Hough transform (2009, methodology): “The choice of local features is arbitrary and we choose local features based on patches of fixed size for simplicity, while taking note that [15] has recently proposed a Hough transform based detector using regions as local features.”
Multiscale Combinatorial Grouping (2014, unknown): “Examples in this family are [18] [13].”
Shape-based pedestrian parsing (2011, methodology): “For modeling parts we use a shape feature built from a spatial histogram of the segment edge orientations (similar to [12]).”
Sketch Tokens: A Learned Mid-level Representation for Contour and Object Detection (2013, background): “Representation based on regions [13] are also popular.”
Bottom-Up Segmentation for Top-Down Detection (2013, background): “Gu et al. [17] apply hough transform for a set of regions to cast votes on the location of the object.”
Object Recognition by Sequential Figure-Ground Ranking (2011, methodology): “Comparisons with (Gu et al. 2009) in Table 3 show improvement in most classes.”
Pedestrian Detection: An Evaluation of the State of the Art (2012, background): “These appear most promising for low to medium resolution settings, under which segmentation [35] or keypoint [36], [37] based methods often fail.”
Oriented Object Proposals (2015, methodology): “Gu et al. [16] first propose to recognize objects using bounding box proposals obtained from hierarchically grouped superpixels.”