End-to-End Training of Hybrid CNN-CRF Models for Stereo
This addresses the problem of accurate stereo depth estimation for computer vision applications, representing an incremental improvement by integrating CNNs and CRFs in a unified framework.
The paper tackles stereo estimation by proposing a hybrid CNN-CRF model trained end-to-end using structured output SVM, achieving strong performance on benchmarks like Middlebury 2014 and Kitti 2015 without post-processing.
We propose a novel and principled hybrid CNN+CRF model for stereo estimation. Our model allows to exploit the advantages of both, convolutional neural networks (CNNs) and conditional random fields (CRFs) in an unified approach. The CNNs compute expressive features for matching and distinctive color edges, which in turn are used to compute the unary and binary costs of the CRF. For inference, we apply a recently proposed highly parallel dual block descent algorithm which only needs a small fixed number of iterations to compute a high-quality approximate minimizer. As the main contribution of the paper, we propose a theoretically sound method based on the structured output support vector machine (SSVM) to train the hybrid CNN+CRF model on large-scale data end-to-end. Our trained models perform very well despite the fact that we are using shallow CNNs and do not apply any kind of post-processing to the final output of the CRF. We evaluate our combined models on challenging stereo benchmarks such as Middlebury 2014 and Kitti 2015 and also investigate the performance of each individual component.