飞桨学习赛：遥感影像地块分割

Who this competition fits

适合已会用 PaddlePaddle 2.0 及以上版本训练深度学习模型、想做遥感图像语义分割的个人开发者。参赛者需要能处理多类别像素级标签，并将建筑、道路、林地等地物的预测结果导出为分割图。

What this competition gives you

There is a baseline you can run as-is — starting from it is usually far faster than building from scratch.

• DataRemote-sensing segmentation training and test datasets
• TutorialBackground task execution tutorial
• BaselineSelected community project by Jhong1996
• BaselineSelected community project by Jhong1996
• BaselineSelected community project by lxastro
• BaselineSelected community project by iterhui
• BaselineSelected community project by 学霸霸Zidane
• BaselineSelected community project by JavaRoom
• BaselineSelected community project by 开着奔驰种地
• BaselineSelected community project by 杨壳壳
• BaselineSelected community project by 伽利略
• BaselineSelected community project by geoyee
• BaselineSelected community project by 小U
• BaselineSelected community project by Alchemist_W
• BaselineSelected community project by 泸沽
• BaselineSelected community project by dreamkwc
• BaselineSelected community project by JackieTse
• BaselineSelected community project by lzzzzzm
• BaselineCommunity project by cv学习者
• BaselineCommunity project by AlwaysGem
• BaselineCommunity project by JavaRoom
• BaselineCommunity project by AIStudio156983
• BaselineCommunity project by Ylllll

From registration to a first submission

1. 01

   PaddlePaddle 2.0+

2. 02

   语义分割模型

3. 03

   遥感影像处理

4. 04

   像素级标注读取

5. 05

   MIoU评估

6. 06

   PNG预测结果打包

Before you commit: 主要难点是让模型适应多地区、多来源、多场景的异构遥感影像，并做好四类像素级分割；标签中的未标注区域还需要在训练和结果处理时正确对待。提交也有严格格式要求：必须生成与测试图片逐一对应的单通道 PNG，并压缩为 result.zip。