Learning What Matters: Combining Probabilistic Models of 2D and 3D Saliency Cues

Potapova, Ekaterina; Zillich, Michael; Vincze, Markus

doi:10.1007/978-3-642-23968-7_14

Cited by 33 publications

(12 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The models (e.g. [32] and [33]) in this category take depth saliency as additional information. This type of models relies on the existence of "depth saliency maps".…”

Section: Dw Depth Information Operation Validationmentioning

confidence: 99%

“…In most situations, depth contrast can also be an efficient indicator of an interesting target. For example, the HVS might consider a region protruding above a flat plane as a potential target [33]; or might consider a hole as a place where a potential target might exist. In our study, Difference of Gaussians (DoG) filter is applied to the depth map for extracting depth contrast.…”

Section: B a Bayesian Approach Of Depth Saliency Map Generationmentioning

confidence: 99%

“…In the proposed model, we adopt a straightforward approach which is the same as the one used in [33] to merge the depth saliency map (SM dep ) with the 2D saliency map (SM 2D ): the final saliency map SM S is equal to the sum of both maps:…”

Section: B Saliency Maps Combinationmentioning

confidence: 99%

See 2 more Smart Citations

Computational Model of Stereoscopic 3D Visual Saliency

Wang

daSilva

Le-Callet

et al. 2013

IEEE Trans. on Image Process.

119

View full text Add to dashboard Cite

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Abstract-Many computational models of visual attention performing well in predicting salient areas of 2D images have been proposed in the literature. The emerging applications of stereoscopic 3D display bring additional depth information affecting the human viewing behavior, and require extensions of the efforts made in 2D visual modeling. In this paper, we propose a new computational model of visual attention for stereoscopic 3D still image. Apart from detecting salient areas based on 2D visual features, the proposed model takes depth as an additional visual dimension. The measure of depth saliency is derived from the eye movement data obtained from an eye-tracking experiment using synthetic stimuli. Two different ways of integrating depth information in the modeling of 3D visual attention are then proposed and examined. For the performance evaluation of 3D visual attention models, we have created an eye-tracking database which contains stereoscopic images of natural content and is publicly available along with this paper. The proposed model gives a good performance, compared to that of state-of-the-art 2D models on 2D images. The results also suggest that a better performance is obtained when depth information is taken into account through the creation of a depth saliency map rather than when it is integrated by a weighting method.

show abstract

“…The models (e.g. [32] and [33]) in this category take depth saliency as additional information. This type of models relies on the existence of "depth saliency maps".…”

Section: Dw Depth Information Operation Validationmentioning

confidence: 99%

Section: B a Bayesian Approach Of Depth Saliency Map Generationmentioning

confidence: 99%

See 1 more Smart Citation

Computational Model of Stereoscopic 3D Visual Saliency

Wang

daSilva

Le-Callet

et al. 2013

IEEE Trans. on Image Process.

119

View full text Add to dashboard Cite

show abstract

“…We have extended the primarily 2D attention cues and exploited stereo and RGB-D images to include 3D pre-attentive cues as also known to be used in humans. The main novelty of the work lies in the understanding how and what pre-attentive cues should be combined for the estimation of attention points useful for segmentation of graspable objects [17].…”

Section: Main Findings and Resultsmentioning

confidence: 99%

Grasping Grasping in GRASP

Kragić

2013

JRSJ

View full text Add to dashboard Cite

Grasping and manipulation of objects is an important research problem in the area of robotics and has been addressed both from an analytical and an empirical perspective. In Europe, there have been several large-scale projects that studied different aspects of this problem: HANDLE, GRASP, DEXMART, PACO-PLUS, TOMSY, THE, to name a few. We summarize some of the main ideas and results from the GRASP project (www.grasp-project.eu).

show abstract

“…4, verwenden wir auch 3D-Hinweise. Die Abbildung zeigt deutlich, dass die 3D-Hinweise eine weitaus bessere Segmentierung der Szene erm€ oglichen (f€ ur Details verweisen wir auf (Potapova, Zillich, Vincze, 2011)). Somit wird die nachfolgende Objektklassifizierung drastisch vereinfacht.…”

Section: Erkennung Horizontaler Ebenenunclassified

Situiertes Sehen für bessere Erkennung von Objekten und Objektklassen

Vincze

Wohlkinger

Aldóma

et al. 2012

Elektrotech. Inftech.

View full text Add to dashboard Cite

Eine Hauptaufgabe f€ ur Roboter ist es, Objekte und Objektklassen zu erkennen, um diese zu finden und handzuhaben. Bild-basierte Ans€ atze lernen aus großen Datenbanken, haben aber keinen Zugriff auf Kontextwissen zur Verf€ ugung, zum Beispiel, wie Roboter in Zimmern navigieren. Wir schlagen daher den Ansatz des situierten Sehens vor, um kontextuelles Wissen € uber die Aufgabe und die Anwendung zur Verbesserung der Objekterkennung zu verwenden. Basierend auf der Bestimmung des freien Bodens vor dem Roboter, der f€ ur die sichere Navigation notwendig ist, zeigen wir, dass dadurch die Lokalisierung, das Erkennung von Fl€ achen und die Kategorisierung von Objekten vereinfacht werden. Wir zeigen ferner, dass Objektklassen effizient aus 3D-Web-Daten gelernt werden k€ onnen, wenn das Lernen virtuelle 2,5D-Ansichten verwendet, um die Sicht der Sensoren des Roboters auf die reale Welt anzunehmen. Mit diesem Ansatz wurden 200 Objektklassen (zu finden unter www.3d-net.org) modelliert und in Szenen erkannt, z. B. St€ uhle mit einer Erkennungsrate von 93 Prozent. Schlü sselwö rter: Objekterkennung; Klassifizierung; Roboter zu Hause; situiertes SehenSituated vision for improved detection of objects and object classes.A main task for domestic robots is to recognize object categories. Image-based approaches learn from large data bases but have no access to contextual knowledge such as available to a robot navigating in the rooms at home. Consequently, we set out to exploit the knowledge available to the robot to constrain the task of object classification. Based on the estimation of free ground space in front of the robot, which is essential for the save navigation in a home setting, we show that we can greatly simplify self-localization, the detection of support surfaces, and the classification of objects. We further show that object classes can be efficiently acquired from 3D models of the Web if learned from automatically generated view data. We modelled 200 object classes (available from www.3d-net.org) and provide sample data of scenes for testing. Using the situated approach we can detect, e.g., chairs with 93 per cent detection rate. EinleitungMenschen tr€ aumen immer wieder von Robotern, die uns zu Hause l€ astige Arbeiten abnehmen. Trotz der Fortschritte in der Robotik ist der Weg dorthin weit. Die gr€ oßten Herausforderungen liegen in der großen Variabilit€ at der Umgebung und der Anzahl der Objekte. Der Roboter muss zwischen vielen verschiedenen M€ obeln und Raumarten sicher navigieren und lokalisieren, und er muss eine sehr große Anzahl an Objekten und Objektklassen kennen. Denn es ist schwierig, dem Roboter Befehle zu erteilen, wie "Bitte hier aufr€ aumen" oder "Bitte den Tisch decken", wenn er nicht alle Objekte kennt, die dazu notwendig sind. W€ ahrend es realistisch erscheint, dass ein zuk€ unftiger Besitzer dem Roboter einige spezielle Objekte zeigt, wird man die Lust am Roboter schnell verlieren, wenn dies f€ ur jeden Gegenstand notwendig ist.Das Ziel dieses Beitrages ist aufzuzeigen, wo der Stand der Technik ...

show abstract

Learning What Matters: Combining Probabilistic Models of 2D and 3D Saliency Cues

Cited by 33 publications

References 21 publications

Computational Model of Stereoscopic 3D Visual Saliency

Computational Model of Stereoscopic 3D Visual Saliency

Grasping Grasping in GRASP

Situiertes Sehen für bessere Erkennung von Objekten und Objektklassen

Contact Info

Product

Resources

About