go>> side-fda>> 返回
项目作者： GabrielMajeri

项目描述 ：
PyTorch reimplementation of the "Single Image Depth Estimation Using Fourier Domain Analysis" paper
高级语言： Python
项目地址: git://github.com/GabrielMajeri/side-fda.git
创建时间： 2019-03-27T05:41:09Z
项目社区：https://github.com/GabrielMajeri/side-fda
开源协议：Mozilla Public License 2.0
下载


PyTorch implementation of “Single-Image Depth Estimation Using Fourier Domain Analysis”

Description

This repository contains an open source implementation of a CVPR 2018 paper
which reconstructs the depth of a scene based on a color image of it.

The paper achieves competitive results by:

  • reusing an existing CNN architecture (ResNet-152) and extending it with some new,
    specialized convolutional layers

  • introducing a new loss function which improves the accuracy of the depth estimation

  • running the network on the image and on smaller cropped sections of the image, then
    recombining the results using Fourier domain analysis

Implementation details

The original implementation, available on the authors’ site,
was written in MATLAB and used Caffe.
This implementation is based on the PyTorch library for doing deep learning in Python.

License

The code in this repository is licensed under the Mozilla Public License Version 2.0,
see the LICENSE file for the full text.