{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"从类别像素走向逐个对象","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.md","pages":[{"id":"reading-opening","title":"从类别像素走向逐个对象","unit":"从类别像素走向逐个对象","unitId":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","chapter":"course","number":1},{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","unit":"从类别像素走向逐个对象","unitId":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","chapter":"course","number":2}],"aliases":[{"oldId":"b-3e1a543aa9e1fcac761606e1","targetId":"b-3e1a543aa9e1fcac761606e1","operation":"historical-explanation","sourceScope":"document"},{"oldId":"b-4f1ecd5d6e93ec240bb6005f","targetId":"b-4f1ecd5d6e93ec240bb6005f","operation":"historical-explanation","sourceScope":"document"},{"oldId":"b-0b37d87480f7f96d87417a2a","targetId":"b-0b37d87480f7f96d87417a2a","operation":"historical-explanation","sourceScope":"document"},{"oldId":"b-98c906086e0978e67816b78b","targetId":"b-98c906086e0978e67816b78b","operation":"historical-explanation","sourceScope":"document"},{"oldId":"b-cec9ee1e2a3338b2b4a5252f","targetId":"b-cec9ee1e2a3338b2b4a5252f","page":"reading-opening","operation":"move","beforeSHA256":"83aaf240c580bb5e1a3c458920748ed89991c349313e733e2c8bb2d5ae6ac25d","afterSHA256":"83aaf240c580bb5e1a3c458920748ed89991c349313e733e2c8bb2d5ae6ac25d","reason":"Same source body in authored semantic page"},{"oldId":"b-6a24704d7b32e5f1cc6a3cad","targetId":"b-6a24704d7b32e5f1cc6a3cad","page":"reading-opening","operation":"move","beforeSHA256":"f64939b87b15e9995d2f293ac5ab4037c86703eca1c0d113c25c355d3358549e","afterSHA256":"f64939b87b15e9995d2f293ac5ab4037c86703eca1c0d113c25c355d3358549e","reason":"Same source body in authored semantic page"},{"oldId":"b-344d12bdb7b8f58b28a184fb","targetId":"b-344d12bdb7b8f58b28a184fb","page":"reading-opening","operation":"move","beforeSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","afterSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","reason":"Same source body in authored semantic page"},{"oldId":"b-99a8480ae87fe5403b0bb1c0","targetId":"b-99a8480ae87fe5403b0bb1c0","page":"reading-opening","operation":"move","beforeSHA256":"70eec56f79b131d4695a1c420bc695a02ef3863727481994de8f7ed4bd673dcc","afterSHA256":"70eec56f79b131d4695a1c420bc695a02ef3863727481994de8f7ed4bd673dcc","reason":"Same source body in authored semantic page"},{"oldId":"b-85fc8d4ba02afc8a7fdff49f","targetId":"b-85fc8d4ba02afc8a7fdff49f","page":"reading-opening","operation":"move","beforeSHA256":"e4fe0386a7ec4009e0b726ed0380c146b413771e54cc9b67fa42f3c17ddc9e69","afterSHA256":"e4fe0386a7ec4009e0b726ed0380c146b413771e54cc9b67fa42f3c17ddc9e69","reason":"Same source body in authored semantic page"},{"oldId":"b-51917d2a58ac97fc848b23c9","targetId":"b-51917d2a58ac97fc848b23c9","page":"reading-opening","operation":"move","beforeSHA256":"62cd3c3e7472a4976a8c101608c68d04df6eb55e6ed8b1cdfe8b808e43846633","afterSHA256":"62cd3c3e7472a4976a8c101608c68d04df6eb55e6ed8b1cdfe8b808e43846633","reason":"Same source body in authored semantic page"},{"oldId":"b-298a5ae3775cf6d066f308f7","targetId":"b-298a5ae3775cf6d066f308f7","page":"reading-opening","operation":"move","beforeSHA256":"b77c69f411f0cec66906efa93953096b72fee0517dd04ceb2eb4f2c14fca2c6c","afterSHA256":"b77c69f411f0cec66906efa93953096b72fee0517dd04ceb2eb4f2c14fca2c6c","reason":"Same source body in authored semantic page"},{"oldId":"b-938f9a37258f4301ebfbbc3f","targetId":"b-938f9a37258f4301ebfbbc3f","page":"reading-opening","operation":"move","beforeSHA256":"df469b42a008cfb7ecd8a88fd7f7836c4bc0779d31028b4938705dc288908ff6","afterSHA256":"df469b42a008cfb7ecd8a88fd7f7836c4bc0779d31028b4938705dc288908ff6","reason":"Same source body in authored semantic page"},{"oldId":"b-df37a39b84b09986cc0f03bc","targetId":"b-df37a39b84b09986cc0f03bc","page":"reading-opening","operation":"move","beforeSHA256":"ea717067584e344418a930c49862e5bea35e54c3e75453a5dc35674aa7112bb5","afterSHA256":"ea717067584e344418a930c49862e5bea35e54c3e75453a5dc35674aa7112bb5","reason":"Same source body in authored semantic page"},{"oldId":"b-c62242c4cf6d6cf71aea671c","targetId":"b-c62242c4cf6d6cf71aea671c","page":"reading-opening","operation":"move","beforeSHA256":"fdde3520bab89e32b78dca4f14b82d41aa943c01e54d2b8a789f56ae0adbb053","afterSHA256":"fdde3520bab89e32b78dca4f14b82d41aa943c01e54d2b8a789f56ae0adbb053","reason":"Same source body in authored semantic page"},{"oldId":"b-70cf4f6546881f1e76990a65","targetId":"b-70cf4f6546881f1e76990a65","page":"reading-section-1","operation":"move","beforeSHA256":"e9fee9ad6cac4094a8f46cb3106e8782d3fee4eecd1c1f2207ca52af17c2dec0","afterSHA256":"e9fee9ad6cac4094a8f46cb3106e8782d3fee4eecd1c1f2207ca52af17c2dec0","reason":"Same source body in authored semantic page"},{"oldId":"b-1b2367e4ad61fe19e0e430c1","targetId":"b-1b2367e4ad61fe19e0e430c1","page":"reading-section-1","operation":"rewrite","beforeSHA256":"eb9ce84712b6c0b764c42fda389bb02c8469bedac54010965c3b3f790e46d4e0","afterSHA256":"e2deae2a61518dc33f64cf5fcf53f9feb4901c2f5cf511ac72edb945dd9ab022","reason":"Same source identity with figure semantics or reading context enriched"},{"oldId":"b-804f105e489097d6b23cacc3","targetId":"b-804f105e489097d6b23cacc3","page":"reading-section-1","operation":"move","beforeSHA256":"b2da5a7d8caaa632c7ae9f9f6ce0a06fe464eecb40e6472dfe0da761208eb634","afterSHA256":"b2da5a7d8caaa632c7ae9f9f6ce0a06fe464eecb40e6472dfe0da761208eb634","reason":"Same source body in authored semantic page"},{"oldId":"b-b329148835f7588030aecba6","targetId":"b-b329148835f7588030aecba6","page":"reading-section-1","operation":"move","beforeSHA256":"4f70352dd5540e712df3eec48a1d8a256a043e13993dd8de8b43b7b9f446b8b8","afterSHA256":"4f70352dd5540e712df3eec48a1d8a256a043e13993dd8de8b43b7b9f446b8b8","reason":"Same source body in authored semantic page"},{"oldId":"b-d8f52be1b5a2abd4f05ef930","targetId":"b-d8f52be1b5a2abd4f05ef930","page":"reading-section-1","operation":"move","beforeSHA256":"acad29e92db22a1ad00a309903470a4328f8c18cfdf4fac20c3446621f0f6335","afterSHA256":"acad29e92db22a1ad00a309903470a4328f8c18cfdf4fac20c3446621f0f6335","reason":"Same source body in authored semantic page"},{"oldId":"b-614bc35b5161bc66cab4982a","targetId":"b-614bc35b5161bc66cab4982a","page":"reading-section-1","operation":"move","beforeSHA256":"6d41a42fe14ef8b81dbbf4c540876335e4e04e4ecf75f45c1c416ef808b7f693","afterSHA256":"6d41a42fe14ef8b81dbbf4c540876335e4e04e4ecf75f45c1c416ef808b7f693","reason":"Same source body in authored semantic page"},{"oldId":"b-644472a62d1396c71c6a0445","targetId":"b-644472a62d1396c71c6a0445","page":"reading-section-1","operation":"move","beforeSHA256":"6ccfb455e3b10fd39f061e5a79711a72b3121ff01414683996e829721f063b92","afterSHA256":"6ccfb455e3b10fd39f061e5a79711a72b3121ff01414683996e829721f063b92","reason":"Same source body in authored semantic page"},{"oldId":"b-b00bbfd8a344f4b2101e061f","targetId":"b-b00bbfd8a344f4b2101e061f","page":"reading-section-1","operation":"move","beforeSHA256":"a27c6258b1428ede911a5cd9e234dac0995bdd9caa159cef345082b9ebf07a51","afterSHA256":"a27c6258b1428ede911a5cd9e234dac0995bdd9caa159cef345082b9ebf07a51","reason":"Same source body in authored semantic page"},{"oldId":"block-8413a612fc314b05806c00e9cf00fb7a--从类别像素走向逐个对象","targetId":"b-85fc8d4ba02afc8a7fdff49f","operation":"retain-source-identity"}],"blocks":[{"id":"b-cec9ee1e2a3338b2b4a5252f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-cec9ee1e2a3338b2b4a5252f","content":"[CLS 08 · 密集预测、实例与集合输出](/course/notebooks/2d81e2a1936043349eacb71c53c0defa.html)","sha256":"83aaf240c580bb5e1a3c458920748ed89991c349313e733e2c8bb2d5ae6ac25d","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-6a24704d7b32e5f1cc6a3cad","prerequisites":[],"relatedBlocks":[]},{"id":"b-6a24704d7b32e5f1cc6a3cad","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-6a24704d7b32e5f1cc6a3cad","content":"# 从类别像素走向逐个对象","sha256":"f64939b87b15e9995d2f293ac5ab4037c86703eca1c0d113c25c355d3358549e","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-cec9ee1e2a3338b2b4a5252f","nextBlock":"b-344d12bdb7b8f58b28a184fb","prerequisites":[{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-344d12bdb7b8f58b28a184fb","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-344d12bdb7b8f58b28a184fb","content":"按问题继续","sha256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6a24704d7b32e5f1cc6a3cad","nextBlock":"b-99a8480ae87fe5403b0bb1c0","prerequisites":[{"id":"b-6a24704d7b32e5f1cc6a3cad","url":"#b-6a24704d7b32e5f1cc6a3cad","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-99a8480ae87fe5403b0bb1c0","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-99a8480ae87fe5403b0bb1c0","content":"像素损失、IoU口径、粗细表示、实例身份与集合输出","sha256":"70eec56f79b131d4695a1c420bc695a02ef3863727481994de8f7ed4bd673dcc","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-344d12bdb7b8f58b28a184fb","nextBlock":"b-85fc8d4ba02afc8a7fdff49f","prerequisites":[{"id":"b-6a24704d7b32e5f1cc6a3cad","url":"#b-6a24704d7b32e5f1cc6a3cad","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-85fc8d4ba02afc8a7fdff49f","type":"h2","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-85fc8d4ba02afc8a7fdff49f","content":"## 从类别像素走向逐个对象","sha256":"e4fe0386a7ec4009e0b726ed0380c146b413771e54cc9b67fa42f3c17ddc9e69","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-99a8480ae87fe5403b0bb1c0","nextBlock":"b-51917d2a58ac97fc848b23c9","prerequisites":[{"id":"b-6a24704d7b32e5f1cc6a3cad","url":"#b-6a24704d7b32e5f1cc6a3cad","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-51917d2a58ac97fc848b23c9","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-51917d2a58ac97fc848b23c9","content":"把像素标为杯子类别，只声明“这里属于杯子集合”。实例标注还声明“这里属于杯一，那里属于杯二”。两只杯子即使同色、接触甚至相互遮挡，实例任务仍需给出两份对象身份。","sha256":"62cd3c3e7472a4976a8c101608c68d04df6eb55e6ed8b1cdfe8b808e43846633","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-85fc8d4ba02afc8a7fdff49f","nextBlock":"b-298a5ae3775cf6d066f308f7","prerequisites":[{"id":"b-85fc8d4ba02afc8a7fdff49f","url":"#b-85fc8d4ba02afc8a7fdff49f","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-298a5ae3775cf6d066f308f7","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-298a5ae3775cf6d066f308f7","content":"考虑一行标签 `[1,1,2,2]`，其中 1、2 是两只杯子的实例号。将它们都映射到类别杯子=1，就得到 `[1,1,1,1]`。另一个实例划分 `[1,2,2,2]` 经同样映射也得到同一类别行。因此类别图不能唯一反推出实例边界。","sha256":"b77c69f411f0cec66906efa93953096b72fee0517dd04ceb2eb4f2c14fca2c6c","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-51917d2a58ac97fc848b23c9","nextBlock":"b-938f9a37258f4301ebfbbc3f","prerequisites":[{"id":"b-85fc8d4ba02afc8a7fdff49f","url":"#b-85fc8d4ba02afc8a7fdff49f","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-938f9a37258f4301ebfbbc3f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-938f9a37258f4301ebfbbc3f","content":"连通域可以在“每个对象互不接触”的特定条件下帮忙计数，但一旦两只杯子相接，一个连通块可含两个对象；反过来同一被遮挡的对象还可能产生不连通的可见部分。对象身份不是连通性规则自动保证的属性。","sha256":"df469b42a008cfb7ecd8a88fd7f7836c4bc0779d31028b4938705dc288908ff6","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-298a5ae3775cf6d066f308f7","nextBlock":"b-df37a39b84b09986cc0f03bc","prerequisites":[{"id":"b-85fc8d4ba02afc8a7fdff49f","url":"#b-85fc8d4ba02afc8a7fdff49f","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-df37a39b84b09986cc0f03bc","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-df37a39b84b09986cc0f03bc","content":"区域路线给每个候选一份身份，像素路线给精细轮廓。把它们结合，就可以在“这个候选对象”之内预测局部 mask。这是进入 Mask R-CNN 的具体需求，而不是为了再增加一种模型名称。","sha256":"ea717067584e344418a930c49862e5bea35e54c3e75453a5dc35674aa7112bb5","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-938f9a37258f4301ebfbbc3f","nextBlock":"b-c62242c4cf6d6cf71aea671c","prerequisites":[{"id":"b-85fc8d4ba02afc8a7fdff49f","url":"#b-85fc8d4ba02afc8a7fdff49f","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-c62242c4cf6d6cf71aea671c","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-c62242c4cf6d6cf71aea671c","content":"[Mask R-CNN 的三项输出与局部目标](/course/blocks/mask-branch-target.html) · [同一杯子换RoI后，目标如何改变](/course/blocks/local-mask-coordinate-target.html) · [P40](/course/pages/l3-r2-p40.html)—[P43](/course/pages/l3-r2-p43.html)。","sha256":"fdde3520bab89e32b78dca4f14b82d41aa943c01e54d2b8a789f56ae0adbb053","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-opening","title":"从类别像素走向逐个对象","number":1},"sectionTitle":"从类别像素走向逐个对象","currentQuestion":"从类别像素走向逐个对象","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-df37a39b84b09986cc0f03bc","nextBlock":null,"prerequisites":[{"id":"b-85fc8d4ba02afc8a7fdff49f","url":"#b-85fc8d4ba02afc8a7fdff49f","role":"section-title"},{"id":"b-cec9ee1e2a3338b2b4a5252f","url":"#b-cec9ee1e2a3338b2b4a5252f","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-446cd304922078174620fb3b","type":"h4","scope":"reading-section-1","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-446cd304922078174620fb3b","content":"#### 选读：不同尺度与无序集合，分别改变哪一部分？","sha256":"86b78c60de3cb89b29e5e15feabf0eec6ebc9dc9cdb49c1a74515ab154cd005f","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-70cf4f6546881f1e76990a65","prerequisites":[{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-70cf4f6546881f1e76990a65","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-70cf4f6546881f1e76990a65","content":"把“有跳连”当成一个共同名字，容易遮住任务差异。FCN 的跳接把较细层的空间线索与粗语义预测结合，服务于逐像素输出；U-Net 的编码—解码路径将高分辨率特征接到相应解码层；FPN 自顶向下融合与横向连接，形成可供不同尺度对象使用的特征层级。","sha256":"e9fee9ad6cac4094a8f46cb3106e8782d3fee4eecd1c1f2207ca52af17c2dec0","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-446cd304922078174620fb3b","nextBlock":"b-1b2367e4ad61fe19e0e430c1","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"}],"relatedBlocks":[]},{"id":"b-1b2367e4ad61fe19e0e430c1","type":"table","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-1b2367e4ad61fe19e0e430c1","content":"| 观察点 | 应在计算图里找什么 |\n| --- | --- |\n| 较细信息从哪里来 | 原图或编码器哪个中间层，而非只看上采样箭头 |\n| 怎么对齐 | 两条支路的步长、尺寸、坐标约定 |\n| 怎样组合 | 相加或拼接，组合后是否再有可学习预测 |\n| 服务哪个目标 | 单张类别图、多尺度框、局部实例 mask |","sha256":"e2deae2a61518dc33f64cf5fcf53f9feb4901c2f5cf511ac72edb945dd9ab022","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"table":{"headers":["观察点","应在计算图里找什么"],"rows":[["观察点","应在计算图里找什么"],["较细信息从哪里来","原图或编码器哪个中间层，而非只看上采样箭头"],["怎么对齐","两条支路的步长、尺寸、坐标约定"],["怎样组合","相加或拼接，组合后是否再有可学习预测"],["服务哪个目标","单张类别图、多尺度框、局部实例 mask"]]},"previousBlock":"b-70cf4f6546881f1e76990a65","nextBlock":"b-804f105e489097d6b23cacc3","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-804f105e489097d6b23cacc3","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-804f105e489097d6b23cacc3","content":"如果原图中 4×4 像素的目标进入 stride16 网格，其连续跨度只有 1/4×1/4 格；这使位置读取困难，但不能仅据小于一格断言信息必全丢，通道仍可能记录响应。更细输入或跨层特征带来新的观测机会，是否改善要在指定任务上检查。","sha256":"b2da5a7d8caaa632c7ae9f9f6ce0a06fe464eecb40e6472dfe0da761208eb634","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-1b2367e4ad61fe19e0e430c1","nextBlock":"b-b329148835f7588030aecba6","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-b329148835f7588030aecba6","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-b329148835f7588030aecba6","content":"保留这份比较，是为学生读到结构名时有判断坐标；本轮主讲不再把它拆为几个平行论文段落。","sha256":"4f70352dd5540e712df3eec48a1d8a256a043e13993dd8de8b43b7b9f446b8b8","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-804f105e489097d6b23cacc3","nextBlock":"b-d8f52be1b5a2abd4f05ef930","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-d8f52be1b5a2abd4f05ef930","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-d8f52be1b5a2abd4f05ef930","content":"大量背景参考框已经很容易分类，却可能因数量占优而贡献大量总损失。focal loss 在基本交叉熵外乘 $(1-p_t)^\\gamma$。取 γ=2，正确类概率 .9 的样本权重 .01，概率 .5 的权重 .25，后者相对贡献更大。这是在调整哪些样本值得继续关注，不直接增加空间信息，也不指定哪项预测负责哪只对象。","sha256":"acad29e92db22a1ad00a309903470a4328f8c18cfdf4fac20c3446621f0f6335","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"(1-p_t)^\\gamma","definitionContext":"选读：不同尺度与无序集合，分别改变哪一部分？"}],"previousBlock":"b-b329148835f7588030aecba6","nextBlock":"b-614bc35b5161bc66cab4982a","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-614bc35b5161bc66cab4982a","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-614bc35b5161bc66cab4982a","content":"对象匹配回答另一件事：当真值是一组无序对象，预测槽位该怎样对应它们？给两预测、两真值的代价矩阵 `[[2,.3],[.4,2]]`，对角总代价 4，交叉 .7，所以选择交叉。交换标注数组顺序不该改变对象集合本身，匹配用于消除这个任意顺序。","sha256":"6d41a42fe14ef8b81dbbf4c540876335e4e04e4ecf75f45c1c416ef808b7f693","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-d8f52be1b5a2abd4f05ef930","nextBlock":"b-644472a62d1396c71c6a0445","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-644472a62d1396c71c6a0445","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-644472a62d1396c71c6a0445","content":"LAB 20 从类别概率、框 L1 与 IoU 实际构造成本，再穷举 2! 分配；它是足够透明的解析练习。这里没有图像编码器学习，也不是完整 DETR。主讲保留对象身份的必要性，这些损失组织细节进入选读即可，不必为保持课程丰富而重新开一章。","sha256":"6ccfb455e3b10fd39f061e5a79711a72b3121ff01414683996e829721f063b92","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-614bc35b5161bc66cab4982a","nextBlock":"b-b00bbfd8a344f4b2101e061f","prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-b00bbfd8a344f4b2101e061f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/8413a612fc314b05806c00e9cf00fb7a.html#b-b00bbfd8a344f4b2101e061f","content":"[LAB 20 尺度与集合短算例](https://deepnote.com/workspace/ise3309jingao-5ba66b0a-4283-4f5a-ae06-3f68d3201e99/project/38755e4e-23f6-4f5f-8c68-8a2e4114b1c1/notebook/9315996660f54e92847770355b648f9e?utm_source=openai&utm_medium=mcp&utm_campaign=openaimcp&utm_content=9315996660f54e92847770355b648f9e&utm_term=get_notebook) 保留实际成本构造；不需要为了理解像素目标先学完 DETR。完整注意力结构留在第五讲，不在这里展开。","sha256":"a27c6258b1428ede911a5cd9e234dac0995bdd9caa159cef345082b9ebf07a51","chapter":{"id":"course","title":"从类别像素走向逐个对象"},"unit":{"id":"notebook-blocks/8413a612fc314b05806c00e9cf00fb7a","title":"从类别像素走向逐个对象"},"page":{"id":"reading-section-1","title":"选读：不同尺度与无序集合，分别改变哪一部分？","number":2},"sectionTitle":"选读：不同尺度与无序集合，分别改变哪一部分？","currentQuestion":"选读：不同尺度与无序集合，分别改变哪一部分？","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-644472a62d1396c71c6a0445","nextBlock":null,"prerequisites":[{"id":"b-446cd304922078174620fb3b","url":"#b-446cd304922078174620fb3b","role":"section-title"},{"id":"b-70cf4f6546881f1e76990a65","url":"#b-70cf4f6546881f1e76990a65","role":"problem-context"}],"relatedBlocks":[]}]}
