Generating Object Segmentation Proposals Using Global and Local Search

By Pekka Rantalankila, Juho Kannala, Esa Rahtu. 2014

175 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

What Makes for Effective Detection Proposals? (2015, methodology): “Most papers on generating detection proposals perform fairly limited evaluations, comparing results using only a subset of metrics, datasets, and competing methods.”
Multiscale Combinatorial Grouping for Image Segmentation and Object Proposal Generation (2015, methodology): “Top: original image, instance-level ground truth from COCO and our multiscale hierarchical segmentation.”
Self-taught object localization with deep networks (2014, methodology): “Objectness methods [2, 6, 30, 4, 3, 16, 36, 25, 21, 26, 32, 14] aim at generating bounding boxes that yield high recall for a high number of candidates, i.”
Recurrent Pixel Embedding for Instance Grouping (2017, methodology): “We compare against the most recent techniques including POISE [39], LPO [52], CPMC [8], GOP [51], SeSe [83], GLS [72], RIGOR [38].”
DeepBox: Learning Objectness with Convolutional Networks (2015, neutral): “[26] 50.1 63.3 Ranta[24] 45.9 57.4 SelSearch[29] 55.9 70.8 DeepBox 57.0 83.9 nize other animals.”
How good are detection proposals, really? (2014, background): “2014[25] R4 10 ?? · ? EdgeBoxes[33] EB 0.”
Oriented Object Proposals (2015, background): “Two works [25] [21] extend this idea, where Rantalankila et al. [25] further considers global consistency, and Manén Manén et al. [21] applies a randomized Prim's algorithm for superpixel merging.”
Convolutional Oriented Boundaries: From Image Segmentation to High-Level Tasks (2016, methodology): “In terms of bounding box proposals, we compare also against SharpMask [39], DeepMask [40], EB [49], RPN [48] MCG and SCG [16], LPO [42], BING [50], SeSe [45], GLS [46], RIGOR [47], Obj [51], and RP [52].”
Superpixels: An evaluation of the state-of-the-art (2016, background): “…vision: tracking [10, 11], stereo and occlusion [12, 13], 3D-reconstruction [14], saliency [15], object detection [16, 17], depth recovery [18] and depth estimation [19], indoor scene understanding [20], scene ﬂow [21], and as basis for convolutional neural networks [22] to name just a few.”
Improving object proposals with multi-thresholding straddling expansion (2015, methodology): “Second, diversification strategies required by most models [6, 8, 15, 18, 19, 20] are commonly computationally expensive.”
Reinforcement Learning for Visual Object Detection (2016, methodology): “…function f c : R → R over the set of image regions R: r * = arg max r∈R f c (r) (1) The set of image regions R can be defined either at the coarse level of bounding boxes or at the finer level of freeform image regions obtained with a state of the art proposal generation method[5] [38] [32].”
BING: Binarized normed gradients for objectness estimation at 300fps (2014, methodology): “We extensively evaluate our method on the challenging PASCAL VOC2007 [27] and Microsoft COCO [56] [26], Objectness [4], GOP [49], LPO [50], Rahtu [66], RandomPrim [61], Rantalankila [67], and SelectiveSearch [75] 7 using publicly available code.”
Boosting Object Proposals: From Pascal to COCO (2015, background): “Number of proposals R ec al l MCG [21] SCG [21] CI [10] CPMC [6] GOP [18] GLS [22] RIGOR [15]”
Deep-learning-based object-level contour detection with CCG and CRF optimization (2017, unknown): “Rantalankila [15] 0.”
Learning to segment under various forms of weak supervision (2015, methodology): “To alleviate this issue we adopt a local search algorithm [26] which merges similar adjacent tiny super-pixels.”
Multiscale Combinatorial Grouping for Image Segmentation and Object Proposal Generation (2017, background): “…dependent on the accuracy and the number of object proposals, an increasing body of research has delved into the problem of their generation [12], [13], [14], • J. Pont-Tuset and F. Marques are with the Department of Signal Theory and Communications, Universitat Politècnica de Catalunya,…”