Scene Segmentation Driven by Deep Learning and Surface Fitting

Minto, Ludovico; Pagnutti, Giampaolo; Zanuttigh, Pietro

doi:10.1007/978-3-319-49409-8_12

Cited by 7 publications

(12 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also compared with our previous works, i.e. the approach of [2] based on normalised cuts, the region merging approach of [3] and the method of [4] that represents the starting point for this work. The method of [3] exploits an iterative merging scheme driven by surface fitting but does not exploit any machine learning clue, so it can be used as a reference to evaluate the impact on the performances due to NURBS surface fitting (i.e.…”

Section: Resultsmentioning

confidence: 99%

“…Hasnat et al [13] 2.29 0.90 Hasnat et al [14] 2.20 0.91 Ren et al [16] 2.35 0.90 Felzenszwalb et al [37] 2.32 0.81 Taylor et al [15] 3.15 0.85 Khan et al [8] 2.42 0.87 Dal Mutto et al [2] 3.09 0.84 Pagnutti et al [3] 2.23 0.88 Minto et al [4] 1.93 0.91 proposed method 1.92 0.91…”

Section: Acknowledgmentmentioning

confidence: 99%

“…The approach was firstly proposed in [4], this extended journal version exploits a more advanced classification algorithm and presents a combined segmentation and semantic labelling approach. In particular, with respect to the conference work a deeper convolutional neural network architecture has been employed and surface curvatures and fitting error have also been used inside the CNN (up to our knowledge, this is the first time this kind of data is used in a deep learning framework).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Segmentation and semantic labelling of RGBD data with convolutional neural networks and surface fitting

Pagnutti

Minto

Zanuttigh

2017

IET Computer Vision

Self Cite

View full text Add to dashboard Cite

We present an approach for segmentation and semantic labeling of RGBD data exploiting together geometrical cues and deep learning techniques. An initial over-segmentation is performed using spectral clustering and a set of NURBS surfaces is fitted on the extracted segments. Then a Convolutional Neural Network (CNN) receives in input color and geometry data together with surface fitting parameters. The network is made of nine convolutional stages followed by a softmax classifier and produces a vector of descriptors for each sample. In the next step an iterative merging algorithm recombines the output of the over-segmentation into larger regions matching the various elements of the scene. The couples of adjacent segments with higher similarity according to the CNN features are candidate to be merged and the surface fitting accuracy is used to detect which couples of segments belong to the same surface. Finally, a set of labeled segments is obtained by combining the segmentation output with the descriptors from the CNN. Experimental results show how the proposed approach outperforms state-of-the-art methods and provides an accurate segmentation and labeling.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Acknowledgmentmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Segmentation and semantic labelling of RGBD data with convolutional neural networks and surface fitting

Pagnutti

Minto

Zanuttigh

2017

IET Computer Vision

Self Cite

View full text Add to dashboard Cite

show abstract

“…Naturally, this problem is a harder one. Although deep neural networks can learn to segment scenes [37,23], or show great potential in learning the fitting function and feature extraction [13], jointly solving the detection and surface fitting, to the best of our knowledge, is an open challenge. Moreover, for the problem at hand, data labeling is notoriously exhaustive.…”

Section: D Scenementioning

confidence: 99%

A Minimalist Approach to Type-Agnostic Detection of Quadrics in Point Clouds

Birdal

Busam

Navab

et al. 2018

2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

View full text Add to dashboard Cite

This paper proposes a segmentation-free, automatic and efficient procedure to detect general geometric quadric forms in point clouds, where clutter and occlusions are inevitable. Our everyday world is dominated by man-made objects which are designed using 3D primitives (such as planes, cones, spheres, cylinders, etc.). These objects are also omnipresent in industrial environments. This gives rise to the possibility of abstracting 3D scenes through primitives, thereby positions these geometric forms as an integral part of perception and high level 3D scene understanding.As opposed to state-of-the-art, where a tailored algorithm treats each primitive type separately, we propose to encapsulate all types in a single robust detection procedure. At the center of our approach lies a closed form 3D quadric fit, operating in both primal & dual spaces and requiring as low as 4 oriented-points. Around this fit, we design a novel, local null-space voting strategy to reduce the 4-point case to 3. Voting is coupled with the famous RANSAC and makes our algorithm orders of magnitude faster than its conventional counterparts. This is the first method capable of performing a generic cross-type multi-object primitive detection in difficult scenes. Results on synthetic and real datasets support the validity of our method.

show abstract

“…We developed an ad-hoc Convolutional Neural Network (CNN) structure for this task and its architecture is shown in Fig. 2 As common in many approaches [13,14,15] the network is made of two main parts, namely a set of convolutional layers followed by a linear classification stage.…”

Section: Deep Network Architecturementioning

confidence: 99%

Deep learning for 3D shape classification from multiple depth maps

Zanuttigh

Minto

2017

2017 IEEE International Conference on Image Processing (ICIP)

Self Cite

View full text Add to dashboard Cite

This paper proposes a novel approach for the classification of 3D shapes exploiting deep learning techniques. The proposed algorithm starts by constructing a set of depth maps by rendering the input 3D shape from different viewpoints. Then the depth maps are fed to a multi-branch Convolutional Neural Network. Each branch of the network takes in input one of the depth maps and produces a classification vector by using 5 convolutional layers of progressively reduced resolution. The various classification vectors are finally fed to a linear classifier that combines the outputs of the various branches and produces the final classification. Experimental results on the Princeton ModelNet database show how the proposed approach allows to obtain a high classification accuracy and outperforms several state-of-the-art approaches.

show abstract

Scene Segmentation Driven by Deep Learning and Surface Fitting

Cited by 7 publications

References 26 publications

Segmentation and semantic labelling of RGBD data with convolutional neural networks and surface fitting

Segmentation and semantic labelling of RGBD data with convolutional neural networks and surface fitting

A Minimalist Approach to Type-Agnostic Detection of Quadrics in Point Clouds

Deep learning for 3D shape classification from multiple depth maps

Contact Info

Product

Resources

About