Confidence measures aim at discriminating unreliable disparities inferred by a stereo vision system from reliable ones. A common and effective strategy adopted by most top-performing approaches consists in combining multiple confidence measures by means of an appropriately trained random-forest classifier. In this paper, we propose a novel approach by training an n-channel convolutional neural network on a set of feature maps, each one encoding the outcome of a single confidence measure. This strategy enables to move the confidence prediction problem from the conventional 1D feature maps domain, adopted by approaches based on random-forests, to a more distinctive 3D domain, going beyond single pixel analysis. This fact, coupled with a deep network appropriately trained on a small subset of images, enables to outperform top-performing approaches based on random-forests.
Even More Confident Predictions with Deep Machine-Learning / Poggi, Matteo; Tosi, Fabio; Mattoccia, Stefano. - ELETTRONICO. - 2017-:(2017), pp. 8014788.393-8014788.401. (Intervento presentato al convegno 12th IEEE Embedded Vision Workshop (EVW2017) held in conjunction with IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017), Honolulu, Hawaii (USA) tenutosi a usa nel 2017) [10.1109/CVPRW.2017.54].
Even More Confident Predictions with Deep Machine-Learning
Poggi, MatteoMembro del Collaboration Group
;Tosi, FabioMembro del Collaboration Group
;Mattoccia, StefanoMembro del Collaboration Group
2017
Abstract
Confidence measures aim at discriminating unreliable disparities inferred by a stereo vision system from reliable ones. A common and effective strategy adopted by most top-performing approaches consists in combining multiple confidence measures by means of an appropriately trained random-forest classifier. In this paper, we propose a novel approach by training an n-channel convolutional neural network on a set of feature maps, each one encoding the outcome of a single confidence measure. This strategy enables to move the confidence prediction problem from the conventional 1D feature maps domain, adopted by approaches based on random-forests, to a more distinctive 3D domain, going beyond single pixel analysis. This fact, coupled with a deep network appropriately trained on a small subset of images, enables to outperform top-performing approaches based on random-forests.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.