{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"LAB20 的两个槽位为何应交叉匹配","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html","markdown":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.md","pages":[{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","unit":"选读：输出组织与训练分配还可以怎样改变？","unitId":"spatial-training-extensions","chapter":"II","number":1}],"aliases":[{"oldId":"b-2f9e1583f9ea2c917acd4337","targetId":"b-2f9e1583f9ea2c917acd4337","page":"reading-opening","operation":"move","beforeSHA256":"61011e34fd1453a9f761d8405a686334f04d8e0dc77934644c911882e99a88cc","afterSHA256":"61011e34fd1453a9f761d8405a686334f04d8e0dc77934644c911882e99a88cc","reason":"Same source body in authored semantic page"},{"oldId":"b-e1e9ec95ff1b8bfabd8d1cad","targetId":"b-e1e9ec95ff1b8bfabd8d1cad","page":"reading-opening","operation":"move","beforeSHA256":"66f71cd1faef17a7a06a83c631491bac5b4f376e49bca8b1f468f3604df1787a","afterSHA256":"66f71cd1faef17a7a06a83c631491bac5b4f376e49bca8b1f468f3604df1787a","reason":"Same source body in authored semantic page"},{"oldId":"b-1d7a9bc6a1d4404a37ab863c","targetId":"b-1d7a9bc6a1d4404a37ab863c","page":"reading-opening","operation":"move","beforeSHA256":"8c4a4fcaec0e6c0eb856799d47828d49dcd70cbeddd2d724b59395207d95348e","afterSHA256":"8c4a4fcaec0e6c0eb856799d47828d49dcd70cbeddd2d724b59395207d95348e","reason":"Same source body in authored semantic page"},{"oldId":"b-592209c8baa6f4b80fbb48f0","targetId":"b-592209c8baa6f4b80fbb48f0","page":"reading-opening","operation":"move","beforeSHA256":"666afb23ae32517dfc3373217aa07120f72ac5e338277cf55cef59ec35e29c96","afterSHA256":"666afb23ae32517dfc3373217aa07120f72ac5e338277cf55cef59ec35e29c96","reason":"Same source body in authored semantic page"},{"oldId":"b-a4b0853542c4037acd026d97","targetId":"b-a4b0853542c4037acd026d97","page":"reading-opening","operation":"move","beforeSHA256":"00cdf3ea105cff573d3b5e30cd194b7c8bb22cc8b37fda235912d4f9ef325925","afterSHA256":"00cdf3ea105cff573d3b5e30cd194b7c8bb22cc8b37fda235912d4f9ef325925","reason":"Same source body in authored semantic page"},{"oldId":"b-2e0a34ad49a4223eb0642b3b","targetId":"b-2e0a34ad49a4223eb0642b3b","page":"reading-opening","operation":"move","beforeSHA256":"2c901bf98cd806f477e314f0897e9351b3172089a6eab29d4f2569837c1837e6","afterSHA256":"2c901bf98cd806f477e314f0897e9351b3172089a6eab29d4f2569837c1837e6","reason":"Same source body in authored semantic page"},{"oldId":"b-c63a02cc96955d1e540cc516","targetId":"b-c63a02cc96955d1e540cc516","page":"reading-opening","operation":"move","beforeSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","afterSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-2f9e1583f9ea2c917acd4337","type":"h1","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-2f9e1583f9ea2c917acd4337","content":"# LAB20 的两个槽位为何应交叉匹配","sha256":"61011e34fd1453a9f761d8405a686334f04d8e0dc77934644c911882e99a88cc","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-e1e9ec95ff1b8bfabd8d1cad","prerequisites":[{"id":"b-e1e9ec95ff1b8bfabd8d1cad","url":"#b-e1e9ec95ff1b8bfabd8d1cad","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-e1e9ec95ff1b8bfabd8d1cad","type":"p","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-e1e9ec95ff1b8bfabd8d1cad","content":"从错类、漏检、框偏和重复四项错误读结果。IoU 把位置偏差变成可比较几何量，AP 再加入分数排序与对象唯一匹配。沿下方积分例走完分数排序、精确率与召回率计算，再用 LAB10 的结果辨认这些错误。","sha256":"66f71cd1faef17a7a06a83c631491bac5b4f376e49bca8b1f468f3604df1787a","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-2f9e1583f9ea2c917acd4337","nextBlock":"b-1d7a9bc6a1d4404a37ab863c","prerequisites":[{"id":"b-2f9e1583f9ea2c917acd4337","url":"#b-2f9e1583f9ea2c917acd4337","role":"section-title"}],"relatedBlocks":[]},{"id":"b-1d7a9bc6a1d4404a37ab863c","type":"p","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-1d7a9bc6a1d4404a37ab863c","content":"当前 LAB20 的第一项预测框在右侧，类别概率更支持第二类；第二项预测在左侧，更支持第一类。真值数组却按左、右保存。如果不考虑内容，直接让槽位 0 学真值 0，就人为制造了位置和类别冲突。","sha256":"8c4a4fcaec0e6c0eb856799d47828d49dcd70cbeddd2d724b59395207d95348e","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-e1e9ec95ff1b8bfabd8d1cad","nextBlock":"b-592209c8baa6f4b80fbb48f0","prerequisites":[{"id":"b-2f9e1583f9ea2c917acd4337","url":"#b-2f9e1583f9ea2c917acd4337","role":"section-title"},{"id":"b-e1e9ec95ff1b8bfabd8d1cad","url":"#b-e1e9ec95ff1b8bfabd8d1cad","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-592209c8baa6f4b80fbb48f0","type":"p","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-592209c8baa6f4b80fbb48f0","content":"代码按教学定义 $C_{ij}=-\\log p_i(c_j)+2\\lVert b_i-g_j\\rVert_1+(1-IoU)$ 构造成本，得到约 `[[5.0971,.4509],[.4779,5.6026]]`。对角总代价 10.6997，交叉 .9288，因此最佳分配为槽位 0→真值 1、槽位 1→真值 0。只交换标注或预测数组的排列，最小内容对应仍然不变。","sha256":"666afb23ae32517dfc3373217aa07120f72ac5e338277cf55cef59ec35e29c96","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"C_{ij}=-\\log p_i(c_j)+2\\lVert b_i-g_j\\rVert_1+(1-IoU)","definitionContext":"LAB20 的两个槽位为何应交叉匹配"}],"previousBlock":"b-1d7a9bc6a1d4404a37ab863c","nextBlock":"b-a4b0853542c4037acd026d97","prerequisites":[{"id":"b-2f9e1583f9ea2c917acd4337","url":"#b-2f9e1583f9ea2c917acd4337","role":"section-title"},{"id":"b-e1e9ec95ff1b8bfabd8d1cad","url":"#b-e1e9ec95ff1b8bfabd8d1cad","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-a4b0853542c4037acd026d97","type":"p","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-a4b0853542c4037acd026d97","content":"此处的负对数分类项、IoU 与系数是教学选择；官方 DETR matcher 使用其明确的分类、L1、GIoU 成本，不应声称公式逐项一致。匹配找到最低代价也不意味着检测正确：即使两项预测都很差，算法仍会选一个较便宜分配。检测质量仍要用保留数据与定位/类别标准评价。","sha256":"00cdf3ea105cff573d3b5e30cd194b7c8bb22cc8b37fda235912d4f9ef325925","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-592209c8baa6f4b80fbb48f0","nextBlock":"b-2e0a34ad49a4223eb0642b3b","prerequisites":[{"id":"b-2f9e1583f9ea2c917acd4337","url":"#b-2f9e1583f9ea2c917acd4337","role":"section-title"},{"id":"b-e1e9ec95ff1b8bfabd8d1cad","url":"#b-e1e9ec95ff1b8bfabd8d1cad","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-2e0a34ad49a4223eb0642b3b","type":"p","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-2e0a34ad49a4223eb0642b3b","content":"本册另一小算例把 16 像素目标映到不同 stride 的连续跨度，属于尺度读取，与集合匹配不同。导航中把二者标为两项可单独读的短练习，学生无需为找一句机制解释穿过全部实验。","sha256":"2c901bf98cd806f477e314f0897e9351b3172089a6eab29d4f2569837c1837e6","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-a4b0853542c4037acd026d97","nextBlock":"b-c63a02cc96955d1e540cc516","prerequisites":[{"id":"b-2f9e1583f9ea2c917acd4337","url":"#b-2f9e1583f9ea2c917acd4337","role":"section-title"},{"id":"b-e1e9ec95ff1b8bfabd8d1cad","url":"#b-e1e9ec95ff1b8bfabd8d1cad","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-c63a02cc96955d1e540cc516","type":"h2","scope":"set-matching-real-inputs","url":"https://codingai-lec03.pages.dev/course/blocks/set-matching-real-inputs.html#b-c63a02cc96955d1e540cc516","content":"## 继续阅读与实验","sha256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"spatial-training-extensions","title":"选读：输出组织与训练分配还可以怎样改变？"},"page":{"id":"reading-opening","title":"LAB20 的两个槽位为何应交叉匹配","number":1},"sectionTitle":"LAB20 的两个槽位为何应交叉匹配","currentQuestion":"LAB20 的两个槽位为何应交叉匹配","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-2e0a34ad49a4223eb0642b3b","nextBlock":null,"prerequisites":[{"id":"b-2f9e1583f9ea2c917acd4337","url":"#b-2f9e1583f9ea2c917acd4337","role":"section-title"},{"id":"b-e1e9ec95ff1b8bfabd8d1cad","url":"#b-e1e9ec95ff1b8bfabd8d1cad","role":"problem-context"}],"relatedBlocks":[]}]}