Oriented edge forests for boundary detection

By Sam Hallman, Charless C. Fowlkes. 2014

179 citations recorded by the source corpus.

We present a simple, efficient model for learning boundary detection based on a random forest classifier. Our approach combines (1) efficient clustering of training examples based on a simple partitioning of the space of local edge orientations and (2) scale-dependent calibration of individual tree output probabilities prior to multiscale combination. The resulting model outperforms published results on the challenging BSDS500 boundary detection benchmark. Further, on large datasets our model requires substantially less memory for training and speeds up training time by a factor of 10 over the structured forest model.

Subfields: Computer Vision and Pattern Recognition, Ecology, Environmental Engineering, Management, Monitoring, Policy and Law

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 Ecology

In Environmental Engineering

In Management, Monitoring, Policy and Law

Citations in the corpus

EDTER: Edge Detection with Transformer (2022, methodology): “[49], gPb+NG [19], SE [15], SE+NG+ [20], OEF [21], SemiContour [69], HED [65], RCF [36], AMH-Net [66], LPCB [12], BDCN [22], and PiDiNet [53].”
Holistically-Nested Edge Detection (2015, neutral): “…∗ - - 1/10 gPb-owt-ucm [1] .726 .757 .696 1/240 Sketch Tokens [22] .727 .746 .780 1 SCG [29] .739 .758 .773 1/280 SE-Var [6] .746 .767 .803 2.5 OEF [12] .749 .772 .817 - DeepNets [19] .738 .759 .758 1 / 5 † N4-Fields [9] .753 .769 .784 1 / 6 † DeepEdge [2] .753 .772 .807 1/ 10 3 † CSCNN [17] .756…”
RotDCF: Decomposition of Convolutional Filters for Rotation-Equivariant Deep Networks (2018, methodology): “While deep convolutional neural networks (CNN) have been widely used in computer vision and image processing applications, they are not designed to handle large group actions like rotations, which degrade the performance of CNN in many tasks [2, 9, 15, 21, 23].”
Efficient Decomposition of Image and Mesh Graphs by Lifted Multicuts (2015, methodology): “Our experiments are based on [20] which is publicly available and outperformed marginally by [15, 21].”
Oriented Response Networks (2017, background): “In Deep Convolution Neural Networks (DCNNs), the inherent properties of convolution and pooling alleviate the effect of local transitions and warps; however, lacking the capability to handle large image rotation limits DCNN’s performance in many visual tasks including object boundary detection [16, ”
Pixel Difference Networks for Efficient Edge Detection (2021, methodology): “Later learning based methods [16, 9] further utilize various gradient information [59, 37, 12, 15] to produce more accurate boundaries.”
Convolutional Oriented Boundaries: From Image Segmentation to High-Level Tasks (2016, background): “What many of the aforementioned methods have in common is that several simple components contribute to increased performance: (i) information at multiple scales [1], [16], [17], [34], (ii) contour orientation [16], [28], [30], [35], and (iii) end-to-end deep learning [1], [2].”
Learning to predict crisp boundaries (2018, methodology): “Method ODS OIS FPS OEF [49] .651 .667 1/2 gPb-UCM [5] .631 .661 1 /360 gPb+NG [50] .687 .716 1 /375 SE [26] .695 .708 5 SE+NG+ [51] .706 .734 1/15 HED-RGB [12] .720 .734 20 † HED-HHA [12] .682 .695 20 † HED-RGB-HHA [12] .746 .761 10 † RCF-RGB [13] .729 .742 20 † RCF-HHA [13] .705 .715 20 † RCF-RGB-H”
Bi-Directional Cascade Network for Perceptual Edge Detection (2019, methodology): “Table 4 shows the comparison of our method with several recent approaches, including gPb-ucm [1], OEF [17], gPb+NG [15], SE+NG+ [16], SE [10], HED [49], RCF [30] and AMHNet [51].”
Learning Deep Representation for Imbalanced Classification (2016, background): “Edge detection: State-of-the-art edge detection methods [1, 2, 10, 16, 20, 27, 33] mostly use engineered gradient features to classify edge patches against non-edges.”
Richer Convolutional Features for Edge Detection (2016, methodology): “OEF [22] and gPb-UCM [2] only use RGB images, while other methods employ both depth and RGB information.”
Learning Deep Structured Multi-Scale Features using Attention-Gated CRFs for Contour Prediction (2017, methodology): “Method ODS OIS AP gPb-ucm [1] .632 .661 .562 OEF [15] .651 .667 – Silberman et al. [33] .658 .661 – SemiContour [46] .680 .700 .690 SE [10] .685 .699 .679 gPb+NG [13] .687 .716 .629 SE+NG+ [14] .710 .723 .738 HED (RGB) [38] .720 .734 .734 HED (HHA) [38] .682 .695 .702 HED (RGB + HHA) [38] .746 .761…”
Weakly Supervised Object Boundaries (2015, background): “SE is at the core of multiple related methods (SCG, MCG, OEF).”
PixelNet: Towards a General Pixel-level Architecture (2016, unknown): “….640 .580 Felz-Hutt [23] .610 .640 .560 gPb-owt-ucm [2] .726 .757 .696 Sketch Tokens [44] .727 .746 .780 SCG [69] .739 .758 .773 PMI [37] .740 .770 .780 SE-Var [17] .746 .767 .803 OEF [30] .749 .772 .817 DeepNets [39] .738 .759 .758 CSCNN [34] .756 .775 .798 HED [71] .782 .804 .833 HED [71] not.”