Constrained parametric min-cuts for automatic object segmentation

By João Carreira, Cristian Sminchisescu. 2010

488 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

Scalable Object Detection Using Deep Neural Networks (2013, methodology): “In order to address this issue a variety of methods were proposed, varying from detector cascades, to using segmentation to suggest a small number of object hypotheses [14, 2, 4].”
Training Region-Based Object Detectors with Online Hard Example Mining (2016, methodology): “R-CNN, in contrast, uses region proposals [1, 2, 3, 4, 6, 10, 18, 32, 34], a method that was made popular by the selective search algorithm [32].”
What Makes for Effective Detection Proposals? (2015, background): “• Rigor † GC [26] is a somewhat improved variant of CPMC that speeds computation considerably by re-using computation across multiple graph-cut problems and us-ing the fast edge detectors from [34], [38].”
Simultaneous Detection and Segmentation (2014, methodology): “The proposals may be used as just bounding boxes [27] or as region proposals [6,1].”
Category-specific object reconstruction from a single image (2014, background): “Finally, building upon the rapid recent progress in recognition modules [2, 11, 17, 20, 34] (object detection, segmentation and pose estimation), we demonstrate that our learnt models are robust when applied “in the wild” enabling fully automatic reconstructions with just images as inputs.”
Unsupervised Semantic Parsing of Video Collections (2015, methodology): “These proposals are generated using the Constrained Parametric Min-Cut (CPMC) [8] algorithm based on both appearance and motion cues.”
What's the Point: Semantic Segmentation with Point Supervision (2015, methodology): “Such priors have been used in segmentation literature for proposing a set of candidate segments [31], selecting image regions to segment [32], as unary potentials in a conditional random field model [20], or during inference [25].”
HyperNet: Towards Accurate Region Proposal Generation and Joint Object Detection (2016, neutral): “Object proposals [5][22][36][4] considerably reduce the computation compared with sliding window methods [10][26] in detection framework.”
Self-taught object localization with deep networks (2014, methodology): “ds per image, respectively. Furthermore, SELSEARCH exploits speedup heuristics, such as merging only adjacent segments (which we could also adopt to make our code much faster). The methods of [4] and [6] take on average 64 and 432 seconds, respectively. 6. Conclusions This work presents self-taught ”
How good are detection proposals, really? (2014, methodology): “CPMC[4] C 250 - ?? ? Endres2010[9] E 100 - ?? ?? Sel.”
segDeepM: Exploiting segmentation and context in deep neural networks for object detection (2015, methodology): “We obtain object segment proposals via the CPMC approach [3], although our approach is independent of this choice.”
Harmony potentials for joint classification and segmentation (2010, result): “We only provide the results of the winners BONN SVM-SEGM [2] and the method most similar to ours BROOKESMSRC [12], which obtained state-of-the-art results on MSRC-21.”
Object cosegmentation (2011, methodology): “The method of [11] was designed for single image segmentation and we select the highest scoring segmentation as the result.”
Efficient Image and Video Co-localization with Frank-Wolfe Algorithm (2014, result): “This is likely because object proposals in the image domain have received a great deal of attention and study [1, 8, 30, 41] compared to video [5, 7].”
Learning to segment under various forms of weak supervision (2015, methodology): “The final set of effective methods are based on object proposal [4, 9, 1, 11, 16], where class-independent segments are generated, and subsequently classified into different categories using features computed on those segments.”
Top-down visual saliency via joint CRF and dictionary learning (2012, unknown): “F”
Weakly Supervised Localization and Learning with Generic Knowledge (2012, unknown): “We note that as an alternative to the objectness measure of [Alexe et al, 2010b], we could also have used the related methods of [Endres and Hoiem, 2010] or [Carreira et al, 2010 where w n , w m are the widths and h n , h m the heights of windows l n , l m .”
Co-localization in Real-World Images (2014, methodology): “We use the measure of objectness [1], but any method that is able to generate a set of candidate regions can be used [5, 32].”
Video Segmentation with Superpixels (2012, background): “[3–9]) and there is common agreement to use multiple similarities based on brightness, color and texture over local image patches to achieve best image segmentation performance.”