<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Selective Mask Propagation for Multi-Object Tracking · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2606.13033/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Mon, 02 Nov 2026 11:16:22 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/2606.13033/feed.xml" rel="self" type="application/rss+xml" /><item><title>Selective Mask Propagation for Multi-Object Tracking — HN submission</title><link>https://news.ycombinator.com/item?id=48906875</link><guid isPermaLink="false">urn:hackerpapers:submission:2606.13033:48906875</guid><pubDate>Tue, 14 Jul 2026 13:51:19 GMT</pubDate><description>&lt;p&gt;Show HN: Run SAM only when your tracker is uncertain – #1 on SportsMOT&lt;/p&gt;&lt;p&gt;This started when I saw a Roboflow tutorial tracking basketball players by giving every player&amp;#x27;s box to SAM2 and propagating masks through the whole clip (https://www.youtube.com/watch?v=yGQb9KkvQ1Q). Identity preservation worked way better than any dedicated tracker I&amp;#x27;d used before but it only ran at 1–2 fps, where both speed and memory was scaling with player count. Most frames don&amp;#x27;t need SAM at all though. A normal tracking-by-detection tracker matches boxes frame to frame by solving an assignment problem on a cost matrix, and for most of a game that matching is completely trivial. It…&lt;/p&gt;&lt;p&gt;2 points; 0 discussion replies on HN. &lt;a href="https://hackerpaper.kaipfister.de/p/2606.13033/"&gt;Paper and source history&lt;/a&gt;. A submission is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item><item><title>Selective Mask Propagation for Multi-Object Tracking — mentioned by lapurita</title><link>https://news.ycombinator.com/item?id=48906575</link><guid isPermaLink="false">urn:hackerpapers:mention:2606.13033:48906575</guid><pubDate>Tue, 14 Jul 2026 13:30:40 GMT</pubDate><description>&lt;p&gt;This started when I saw a Roboflow tutorial tracking basketball players by giving every player&amp;#x27;s box to SAM2 and propagating masks through the whole clip (https://www.youtube.com/watch?v=yGQb9KkvQ1Q). Identity preservation worked way better than any dedicated tracker I&amp;#x27;d used before but it only ran at 1–2 fps, where both speed and memory was scaling with player count. Most frames don&amp;#x27;t need SAM at all though. A normal tracking-by-detection tracker matches boxes frame to frame by solving an assignment problem on a cost matrix, and for most of a game that matching is completely trivial. It…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2606.13033/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item></channel></rss>
