<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Map on DeepLearning.Earth</title><link>https://deeplearning.earth/tags/map/</link><description>Recent content in Map on DeepLearning.Earth</description><generator>Hugo -- gohugo.io</generator><language>en</language><lastBuildDate>Thu, 22 Oct 2026 06:00:00 +0700</lastBuildDate><atom:link href="https://deeplearning.earth/tags/map/index.xml" rel="self" type="application/rss+xml"/><item><title>Oriented mAP: what to report, and how to read it</title><link>https://deeplearning.earth/posts/2026-10-22_oriented_map/</link><pubDate>Thu, 22 Oct 2026 06:00:00 +0700</pubDate><guid>https://deeplearning.earth/posts/2026-10-22_oriented_map/</guid><description>AP50 asks one question: do the rotated rectangles overlap by half?
That is the number I quote for a DOTA pretrain, and it is the official Task 1 score. On the Hub the field is eval_task1_map50.
Say which images the number was measured on, then give the percentage. How a rotated box becomes that percentage is below.
AP50 counts a match when the rotated rectangles overlap by half. Here is how that percentage is built.</description></item></channel></rss>