文章：
IEEE 3D Vision 2016的文章 原文
Contribution
First, we introduce a fully convolutional architecture to depth prediction, endowed with novel up-sampling blocks, that allows for dense output maps of higher resolution and at the same time requires fewer parameters and trains on one order of magnitude fewer data than the state of the art, while outperforming all existing methods on standard benchmark datasets. We further propose a more efficient scheme for upconvolutions and combine it with the concept of residual learning [7] to create up projection blocks for the effective upsampling of feature maps. Last, we train the network by optimizing a loss based on the reverse Huber function (berHu) [40] and demonstrate, both theoretically and experimentally, why it is beneficial and better suited for the task at hand.
代码:
官方的并没有给出所有源码，给出了tensorflow和matlab版本的模型以及测试方法https://github.com/iro-cp/FCRN-DepthPrediction
我结合网上资源实现用pytorch实现了论文https://github.com/gentlemanman/fcrn_pytorch
但并没有达到论文中的效果，可以试着调整参数继续训练和优化