飞桨学习赛：文档检测优化赛

Who this competition fits

适合已有 Python 与计算机视觉基础、愿意使用 PaddlePaddle 训练端到端模型的个人开发者。尤其适合做过边缘检测、图像分割或关键点/角点回归，并能处理文档照片的候选人；比赛仅支持个人参赛。

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.

• BaselineForkable competition baseline notebook
• TutorialVideo tutorial with score-improvement strategies
• DataDocument-image dataset: 2,797 training and 597 test samples
• Reference paperPixel Difference Networks for Efficient Edge Detection paper
• Reference paperHolistically-Nested Edge Detection paper
• BaselinePiDiNet open-source code reference
• BaselineHEDNet open-source code reference

From registration to a first submission

1. 01

   PaddlePaddle 2.0+

2. 02

   文档边缘检测

3. 03

   图像分割与热力图

4. 04

   四角点坐标回归

5. 05

   OpenCV 图像处理

6. 06

   Python 推理脚本与 ZIP 打包

Before you commit: 难点不只是识别文档边缘，还要在真实拍摄背景下稳定定位文档四个角，并兼顾泛化效果与推理速度。若想取得更高分，页面明确建议直接回归四个角点；同时必须按指定的 predict.py、结果目录和 ZIP 结构提交可运行的模型。