Skip to content
/ ResUnet-a Public

For the semantic segmentation of remote sensing image, tensorflow implementation

70 stars 46 forks Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

ChienWong/ResUnet-a

BranchesTags

Repository files navigation

ResUnet-a

针对遥感影像的语义分割

English introduction here

论文地址[https://arxiv.org/abs/1904.00592]

原作者使用MXNet的实现[https://github.com/feevos/resuneta]

依赖

keras==2.2.2

tensorflow==1.9.0

cv2

numpy

训练

修改train.py中unet.train的第一个参数为数据集地址，第二个参数为模型存放地址

数据集文件格式见[https://github.com/mohuazheliu/ResUnet-a/blob/master/dataset-postdam/train/README.md]

Predict

参照test.py 使用model.predict对一张图片进行预测，使用model.visual可视化预测

Postdam数据集地址

这边将图片压缩一倍后进行训练，训练损失和验证损失

train accuracy train_loss val_accuracy val_loss

测试结果

11

预训练模型下载

OneDrive-Postdam

Paris数据集下载

训练及验证损失

train accuracy train_loss val_accuracy val_loss

测试结果

11

预训练模型下载

OneDrive-Paris

About

For the semantic segmentation of remote sensing image, tensorflow implementation

Resources

Readme
Activity

Stars

70 stars

Watchers

3 watching

Forks

46 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages