<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>FCOS on 111qqz's blog</title><link>https://111qqz.com/en/tags/fcos/</link><description>Recent content in FCOS on 111qqz's blog</description><generator>Hugo -- gohugo.io</generator><language>en</language><managingEditor>hust.111qqz@gmail.com (111qqz)</managingEditor><webMaster>hust.111qqz@gmail.com (111qqz)</webMaster><copyright>© 2011-2026 111qqz</copyright><lastBuildDate>Sun, 20 Sep 2026 13:00:00 +0800</lastBuildDate><atom:link href="https://111qqz.com/en/tags/fcos/index.xml" rel="self" type="application/rss+xml"/><item><title>从 Faster R-CNN 到 Anchor-Free：经典目标检测架构到底在解决什么问题</title><link>https://111qqz.com/en/post/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0/%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A7%86%E8%A7%89/2026-09-20-classical-object-detection-evolution/</link><pubDate>Sun, 20 Sep 2026 13:00:00 +0800</pubDate><author>hust.111qqz@gmail.com (111qqz)</author><guid>https://111qqz.com/en/post/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0/%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A7%86%E8%A7%89/2026-09-20-classical-object-detection-evolution/</guid><description>&lt;p&gt;以前在商汤做 CV 的时候，Faster R-CNN、SSD、FPN、RetinaNet 这些名字是天天打交道的。模型训练、转换、部署，各种坑也踩了不少——&lt;a href="https://111qqz.com/en/post/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0/%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A7%86%E8%A7%89/2019-11-07-debug-frcnn-model-when-upgrading-from-tensorrt4-to-tensorrt5/" &gt;TensorRT 版本间 softmax axis 默认值不一致&lt;/a&gt;、&lt;a href="https://111qqz.com/en/post/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0/%E8%AE%A1%E7%AE%97%E6%9C%BA%E8%A7%86%E8%A7%89/2019-12-13-debug-faster-rcnn-once-again/" &gt;RoI Align 的 num_roi_per_image 默认 300 导致多 batch 结果错乱&lt;/a&gt;，这类问题当年 debug 到怀疑人生。&lt;/p&gt;</description></item></channel></rss>