<?xml version="1.0" encoding="utf-8"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/">
<channel>
<title>滑铁卢华人, 水城百事论坛</title>
<link>http://www.kwcg.ca/forum/</link>
<description>水城社区-Kitchener, Waterloo, Cambridge &amp; Guelph华人的网上家园</description>
<language>en</language>
<item>
<title>四川的乡巴佬自己退下了，哈哈哈哈哈</title>
<content:encoded><![CDATA[<p><em>Posting by 匿名, Thursday, October 08, 2026, 00:29:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551271</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551271</guid>
<pubDate>Thu, 08 Oct 2026 00:29:00 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>华为发布的半年度财报显示，上半年净利润减少近四成，华为常务董事余承东再提“活下来”</title>
<content:encoded><![CDATA[<p><em>Posting by 匿名, Wednesday, October 07, 2026, 23:25:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551270</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551270</guid>
<pubDate>Wed, 07 Oct 2026 23:25:28 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>传说中的核心哥？</title>
<content:encoded><![CDATA[<p><em>Reply by :-D, Wednesday, October 07, 2026, 23:14:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551267</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551267</guid>
<pubDate>Wed, 07 Oct 2026 23:14:23 GMT</pubDate>
<dc:creator>:-D</dc:creator>
</item>
<item>
<title>孙一定是抄了你家祖坟了吧？否则真的没法解释楼猪的龌蹉:-D</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 23:02:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551262</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551262</guid>
<pubDate>Wed, 07 Oct 2026 23:02:42 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>犹太人拿了230多次诺贝尔奖，牛了X了</title>
<content:encoded><![CDATA[<p><em>Posting by 匿名, Wednesday, October 07, 2026, 22:50:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551261</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551261</guid>
<pubDate>Wed, 07 Oct 2026 22:50:06 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>:-D</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名:174.4.*, Wednesday, October 07, 2026, 22:47:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551259</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551259</guid>
<pubDate>Wed, 07 Oct 2026 22:47:41 GMT</pubDate>
<dc:creator>匿名:174.4.*</dc:creator>
</item>
<item>
<title>:-|</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名:174.4.*, Wednesday, October 07, 2026, 22:47:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551258</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551258</guid>
<pubDate>Wed, 07 Oct 2026 22:47:01 GMT</pubDate>
<dc:creator>匿名:174.4.*</dc:creator>
</item>
<item>
<title>你这么想说明你对马斯克的招一无所知。特斯拉会在每辆车里开启FSD，它们没有激活的时候在后台学习司机如何驾驶，如果FSD的判断和司机的判断一致，这个数据被判别为垃圾数据直接丢弃，如果遇到危险的时候人类司机处理方式和FSD不同，并且人类司机安全地避开了事故，这前后10几秒的数据会被传回特斯拉的服务器，作为后期训练的黄金素材。</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 22:44:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551257</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551257</guid>
<pubDate>Wed, 07 Oct 2026 22:44:00 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>骨干</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 22:38:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551256</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551256</guid>
<pubDate>Wed, 07 Oct 2026 22:38:43 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>吐槽一下，这个破班真是上够了，天天要出新方案，还要论证预期效果。</title>
<content:encoded><![CDATA[<p><em>Posting by 匿名, Wednesday, October 07, 2026, 22:28:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551255</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551255</guid>
<pubDate>Wed, 07 Oct 2026 22:28:44 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>送给你吧，谁整天看那张貌像60岁的脸都会厌恶，昨天打泰国选手我们一直都为泰国选手加油的，也不为别的就是那张脸就让人恶心的了。</title>
<content:encoded><![CDATA[<p><em>Reply by 丑丑的乒乓球孙颖莎啊, Wednesday, October 07, 2026, 21:44:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551254</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551254</guid>
<pubDate>Wed, 07 Oct 2026 21:44:33 GMT</pubDate>
<dc:creator>丑丑的乒乓球孙颖莎啊</dc:creator>
</item>
<item>
<title>她得了冠军你娶她可好？:-D :-D</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 21:39:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551253</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551253</guid>
<pubDate>Wed, 07 Oct 2026 21:39:45 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>下届奥运会拿不了女单冠军，它就彻底完蛋了。如果它拿了女单个人冠军那还能和那些老前辈站在同一起点上，不然它就得退到第二档第三档次。</title>
<content:encoded><![CDATA[<p><em>Reply by 丑丑的乒乓球孙颖莎啊, Wednesday, October 07, 2026, 21:36:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551252</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551252</guid>
<pubDate>Wed, 07 Oct 2026 21:36:57 GMT</pubDate>
<dc:creator>丑丑的乒乓球孙颖莎啊</dc:creator>
</item>
<item>
<title>中国乒乓球队史上最无能，但是最会搞营销炒作的球员孙颖莎名声不好，主要是它连奥运会女单都没拿过，人长的丑随它爹，不随我妹妹。</title>
<content:encoded><![CDATA[<p><em>Posting by 匿名, Wednesday, October 07, 2026, 21:31:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551251</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551251</guid>
<pubDate>Wed, 07 Oct 2026 21:31:52 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>臭名昭著的孙丑鬼，我更喜欢她的妈妈不喜欢看她</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 21:29:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551250</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551250</guid>
<pubDate>Wed, 07 Oct 2026 21:29:02 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>更致命的是，特斯拉采用盲训练方法，上亿的参数不知道物理意义是什么，出一次事故找不到需要调整哪些参数，不知道怎样避免同类事故，在重新大量盲训练。</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 21:11:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551249</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551249</guid>
<pubDate>Wed, 07 Oct 2026 21:11:50 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>本来使用lidar “time of fly&quot;的简单问题被搞成这么复杂，可靠性还难以验证，检测时间也拉长，难道不是舍本求末吗？为了省几块钱。</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 21:07:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551248</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551248</guid>
<pubDate>Wed, 07 Oct 2026 21:07:18 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>3. 训练目标变化：增加物理一致性约束。
V13 网络训练更强调空间物理合理性，不只是识别图片里像什么；会判断 “这个空间能不能穿透、是否存在实体阻挡”，而不是单纯依赖车道 / 道路语义特征。
所以即便画面画了逼真的路，只要体素重建判断前方是实心墙，就会判定不可通行，提前减速刹停。</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 20:30:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551247</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551247</guid>
<pubDate>Wed, 07 Oct 2026 20:30:24 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>2. 时序 Transformer + 10–15 秒长期空间记忆（Temporal Persistence）V13 不再只看孤立快照。车辆一边前进，持续用连续多帧视频不断重建、校验三维空间：
真实三维物体，随着车靠近，轮廓、视差、遮挡关系会按物理规律变化；
2D 平面壁画：无论车辆怎么靠近，图案本身不会产生真实 3D 物体的视差变化，时序校验直接识别这是平面贴图幻觉。</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 20:30:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551246</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551246</guid>
<pubDate>Wed, 07 Oct 2026 20:30:13 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
<item>
<title>算法层面的升级才是最重要的部分。
1. Occupancy Network 3.0，体素分辨率提升 8 倍
不再只是识别 “车道、墙、车” 这类物体标签；它把空间切成大量细小 3D 立方体（voxel），预测每一个空间格子是不是被实体物质占用。

- 壁画区域：图像看着像门洞，但体素重建发现门洞图案后面依然是连续实心墙体；
- 模型看到：这个 “门洞” 只是画在障碍物表面的纹理，不是空间里挖空的通道。</title>
<content:encoded><![CDATA[<p><em>Reply by 匿名, Wednesday, October 07, 2026, 20:29:</em></p><p>[ No text ]</p>]]></content:encoded>
<link>http://www.kwcg.ca/forum/index.php?id=1551245</link>
<guid>http://www.kwcg.ca/forum/index.php?id=1551245</guid>
<pubDate>Wed, 07 Oct 2026 20:29:42 GMT</pubDate>
<dc:creator>匿名</dc:creator>
</item>
</channel>
</rss>