Skip to content

This repo is the pytorch implementation of cropland extraction.

Notifications You must be signed in to change notification settings

SonwYang/plough_extraction

Repository files navigation

This repo is the unofficial pytorch implementation of paper "Contour-oriented Cropland Extraction from High Resolution Remote Sensing Imagery Using Richer Convolution Features Network".And I add some tricks on this method.

image-20210624203834465

1. 数据集准备

1.1. 数据制作

  • 利用arcgis制作耕地面矢量;
  • 面矢量转线矢量;
  • 面矢量转栅格,线矢量转栅格;
  • 耕地线膨胀两个像素后,叠加到耕地面上,其中背景的像素值为0,耕地面的像素值为1,耕地线为2。

1.2. 数据目录

data

train_images

  	  ***.tif

  	  ***.tif

  	  ***.tif

         ...........

train_labels

  	  ***.tif

  	  ***.tif

  	  ***.tif

         ...........

2. 模型训练

准备好数据集后, 根据需要进行配置文件的修改,依次训练分割和边缘检测模型,最后进行预测。

3. 项目文件说明

分割模型训练

python train_seg.py

边缘检测模型训练

python train_edge.py

模型推理(分割+边缘+后处理)

python predict_large.py

About

This repo is the pytorch implementation of cropland extraction.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages