Schmähling Tobias, Müller Tobias, Eberhardt Jörg, Elser Stefan
Institute for Photonic Systems Hochschule Ravensburg-Weingarten, University of Applied Sciences, Doggenriedstraße, 88250 Weingarten, Germany.
Institute for Artificial Intelligence Hochschule Ravensburg-Weingarten, University of Applied Sciences, Doggenriedstraße, 88250 Weingarten, Germany.
J Imaging. 2024 Aug 14;10(8):198. doi: 10.3390/jimaging10080198.
In this paper, we present a multi-task model that predicts disparities and confidence levels in deep stereo matching simultaneously. We do this by combining its successful model for each separate task and obtaining a multi-task model that can be trained with a proposed loss function. We show the advantages of this model compared to training and predicting disparity and confidence sequentially. This method enables an improvement of 15% to 30% in the area under the curve (AUC) metric when trained in parallel rather than sequentially. In addition, the effect of weighting the components in the loss function on the stereo and confidence performance is investigated. By improving the confidence estimate, the practicality of stereo estimators for creating distance images is increased.
在本文中，我们提出了一种多任务模型，该模型能够同时预测深度立体匹配中的视差和置信度水平。我们通过结合每个单独任务的成功模型并获得一个可以使用所提出的损失函数进行训练的多任务模型来实现这一点。我们展示了该模型相较于顺序训练和预测视差与置信度的优势。当并行训练而非顺序训练时，此方法在曲线下面积（AUC）指标上能够实现15%至30%的提升。此外，还研究了损失函数中各分量加权对视差和置信度性能的影响。通过改进置信度估计，提高了立体估计器创建距离图像的实用性。
