{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"FCN 直接学习一张类别地图","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html","markdown":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.md","pages":[{"id":"reading-opening","title":"FCN 直接学习一张类别地图","unit":"怎样直接学习一张像素类别地图？","unitId":"pixel-prediction-objective","chapter":"II","number":1}],"aliases":[{"oldId":"b-34f3d5c535b72ca62dbf1297","targetId":"b-34f3d5c535b72ca62dbf1297","page":"reading-opening","operation":"move","beforeSHA256":"18caaee04787a4b6f7f1eae6778d08efbfebdb0443331f1a5c342d4aceb55b65","afterSHA256":"18caaee04787a4b6f7f1eae6778d08efbfebdb0443331f1a5c342d4aceb55b65","reason":"Same source body in authored semantic page"},{"oldId":"b-e59934e7c0600fc92f14a182","targetId":"b-e59934e7c0600fc92f14a182","page":"reading-opening","operation":"move","beforeSHA256":"183d41f9f2dd4315d6651bd668ce69b215479fec798b3e287bba7d446b7f4a31","afterSHA256":"183d41f9f2dd4315d6651bd668ce69b215479fec798b3e287bba7d446b7f4a31","reason":"Same source body in authored semantic page"},{"oldId":"b-3e95a9037a3ed82819a080ec","targetId":"b-3e95a9037a3ed82819a080ec","page":"reading-opening","operation":"move","beforeSHA256":"d764e90ca736ec69dae41ada1a8c4870b608d46d7dc46e61030c9bb13e328e9a","afterSHA256":"d764e90ca736ec69dae41ada1a8c4870b608d46d7dc46e61030c9bb13e328e9a","reason":"Same source body in authored semantic page"},{"oldId":"b-1a0344aa3e9464beb71f0e99","targetId":"b-1a0344aa3e9464beb71f0e99","page":"reading-opening","operation":"move","beforeSHA256":"40e51df4c794999a1c2b86aeb86dcbfba93bcbee0310637690744fac313d48c5","afterSHA256":"40e51df4c794999a1c2b86aeb86dcbfba93bcbee0310637690744fac313d48c5","reason":"Same source body in authored semantic page"},{"oldId":"b-8518631ac64072dfc166972f","targetId":"b-8518631ac64072dfc166972f","page":"reading-opening","operation":"move","beforeSHA256":"d0d7a14e2a82603ebe2f1ef958d50bde08aa07a6ffbf2a378807212760aa648f","afterSHA256":"d0d7a14e2a82603ebe2f1ef958d50bde08aa07a6ffbf2a378807212760aa648f","reason":"Same source body in authored semantic page"},{"oldId":"b-803e122ce31ca0b0bfe4531d","targetId":"b-803e122ce31ca0b0bfe4531d","page":"reading-opening","operation":"move","beforeSHA256":"51e5ab7e0ec16c70833117a3ce12a057283cc6bde9135affeb28a46d5509b290","afterSHA256":"51e5ab7e0ec16c70833117a3ce12a057283cc6bde9135affeb28a46d5509b290","reason":"Same source body in authored semantic page"},{"oldId":"b-94f3e6cb2e8f444b806d5ffd","targetId":"b-94f3e6cb2e8f444b806d5ffd","page":"reading-opening","operation":"rewrite","beforeSHA256":"d2354f4b6fbc87ff038bce996d6bf3cabafca646db38377fa316dda0f22d9e55","afterSHA256":"b6bac54dd1d99034c1142ceb1f6160557738a003215b8f5371883d4a7fced608","reason":"Same source identity with figure semantics or reading context enriched"},{"oldId":"b-c4607cbffdea3a787f657448","targetId":"b-c4607cbffdea3a787f657448","page":"reading-opening","operation":"move","beforeSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","afterSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-34f3d5c535b72ca62dbf1297","type":"h1","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-34f3d5c535b72ca62dbf1297","content":"# FCN 直接学习一张类别地图","sha256":"18caaee04787a4b6f7f1eae6778d08efbfebdb0443331f1a5c342d4aceb55b65","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-e59934e7c0600fc92f14a182","prerequisites":[{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]},{"id":"b-e59934e7c0600fc92f14a182","type":"p","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-e59934e7c0600fc92f14a182","content":"FCN 从逐像素类别图出发，独立于区域检测路线。将 `[B,C,H,W]` logits 与 `[B,H,W]` 标签一一对应，理解分割损失和 ignore 像素。LAB11 从图像预测 mask；已知 mask 的缩放练习只检验采样，不检验视觉预测。","sha256":"183d41f9f2dd4315d6651bd668ce69b215479fec798b3e287bba7d446b7f4a31","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-34f3d5c535b72ca62dbf1297","nextBlock":"b-3e95a9037a3ed82819a080ec","prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]},{"id":"b-3e95a9037a3ed82819a080ec","type":"p","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-3e95a9037a3ed82819a080ec","content":"如果用途是给组织切片的每个像素指定类别，先列一组外接框可能并不是自然起点。FCN 从空间网格出发：把分类网络改成可对不同输入大小输出空间分数图的计算，接受逐像素标签训练，再恢复到目标分辨率。","sha256":"d764e90ca736ec69dae41ada1a8c4870b608d46d7dc46e61030c9bb13e328e9a","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-e59934e7c0600fc92f14a182","nextBlock":"b-1a0344aa3e9464beb71f0e99","prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"},{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]},{"id":"b-1a0344aa3e9464beb71f0e99","type":"p","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-1a0344aa3e9464beb71f0e99","content":"一个 C 类分割模型可以交付 `[B,C,H,W]` logits，每个像素有 C 个实数；目标 `[B,H,W]` 保存该像素的整数类别。类别维 softmax 决定“同一位置是哪类”，空间维则保留“这个判断在哪里”。如果把所有空间位置先平均成一个类别向量，就改了任务。","sha256":"40e51df4c794999a1c2b86aeb86dcbfba93bcbee0310637690744fac313d48c5","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-3e95a9037a3ed82819a080ec","nextBlock":"b-8518631ac64072dfc166972f","prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"},{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]},{"id":"b-8518631ac64072dfc166972f","type":"p","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-8518631ac64072dfc166972f","content":"FCN 与区域检测是两条并行问题线：前者组织像素类别，后者组织对象集合。两条线都能利用分类预训练，却不能把 FCN 描述为 Faster R-CNN 的后继版本。原始 2014/2015 FCN 摘要报告 VOC 2012 的 mean IU 为 62.2%；2016 扩展版本有不同数字，引用时需区分版本。","sha256":"d0d7a14e2a82603ebe2f1ef958d50bde08aa07a6ffbf2a378807212760aa648f","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-1a0344aa3e9464beb71f0e99","nextBlock":"b-803e122ce31ca0b0bfe4531d","prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"},{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]},{"id":"b-803e122ce31ca0b0bfe4531d","type":"p","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-803e122ce31ca0b0bfe4531d","content":"最值得带回科研任务的是这个接口改变：表示能迁移，并不意味着原输出头能直接交付新答案。标注的单位变成像素，损失与评价也须随之改变。","sha256":"51e5ab7e0ec16c70833117a3ce12a057283cc6bde9135affeb28a46d5509b290","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-8518631ac64072dfc166972f","nextBlock":"b-94f3e6cb2e8f444b806d5ffd","prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"},{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]},{"id":"b-94f3e6cb2e8f444b806d5ffd","type":"figure","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-94f3e6cb2e8f444b806d5ffd","content":"图内标签与数值：Lecture 03 课堂图解 P36；区域路线与像素路线，目标不同；R-CNN → Fast → Faster；一组对象：类别 + 框；FCN · 每位置类别分数；语义类别地图；Long · Shelhamer · Darrell，2014 预印本 / CVPR 2015；FCN 是密集像素预测路线，不是 Faster R-CNN 的下一代。\n\n整图的空间特征直接产生逐位置类别分数。 查看对应图解状态\n\n- 区域路线与像素路线，目标不同\n- R-CNN → Fast → Faster\n- 一组对象：类别 + 框\n- FCN · 每位置类别分数\n- 语义类别地图\n- Long · Shelhamer · Darrell，2014 预印本 / CVPR 2015\n- FCN 是密集像素预测路线，不是 Faster R-CNN 的下一代。整图的空间特征直接产生逐位置类别分数。 [查看对应图解状态](/#/l3-r2-p36/0)[查看完整图解](/#/l3-r2-p36/0)","sha256":"b6bac54dd1d99034c1142ceb1f6160557738a003215b8f5371883d4a7fced608","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"figure":{"caption":"整图的空间特征直接产生逐位置类别分数。 查看对应图解状态","labels":["区域路线与像素路线，目标不同","R-CNN → Fast → Faster","一组对象：类别 + 框","FCN · 每位置类别分数","语义类别地图","Long · Shelhamer · Darrell，2014 预印本 / CVPR 2015","FCN 是密集像素预测路线，不是 Faster R-CNN 的下一代。"],"description":"整图的空间特征直接产生逐位置类别分数。 查看对应图解状态区域路线与像素路线，目标不同R-CNN → Fast → Faster一组对象：类别 + 框FCN · 每位置类别分数语义类别地图Long · Shelhamer · Darrell，2014 预印本 / CVPR 2015FCN 是密集像素预测路线，不是 Faster R-CNN 的下一代。","evidenceIdentity":"整图的空间特征直接产生逐位置类别分数。 查看对应图解状态","images":[{"src":"/assets/display/8ced939a1fd55047fbf73ea5.webp","alt":""},{"src":"/assets/display/8ced939a1fd55047fbf73ea5.webp","alt":""}]},"previousBlock":"b-803e122ce31ca0b0bfe4531d","nextBlock":"b-c4607cbffdea3a787f657448","prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"},{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-c4607cbffdea3a787f657448","type":"h2","scope":"fcn-independent-route","url":"https://codingai-lec03.pages.dev/course/blocks/fcn-independent-route.html#b-c4607cbffdea3a787f657448","content":"## 继续阅读与实验","sha256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"pixel-prediction-objective","title":"怎样直接学习一张像素类别地图？"},"page":{"id":"reading-opening","title":"FCN 直接学习一张类别地图","number":1},"sectionTitle":"FCN 直接学习一张类别地图","currentQuestion":"FCN 直接学习一张类别地图","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-94f3e6cb2e8f444b806d5ffd","nextBlock":null,"prerequisites":[{"id":"b-34f3d5c535b72ca62dbf1297","url":"#b-34f3d5c535b72ca62dbf1297","role":"section-title"},{"id":"b-e59934e7c0600fc92f14a182","url":"#b-e59934e7c0600fc92f14a182","role":"problem-context"}],"relatedBlocks":[{"id":"b-94f3e6cb2e8f444b806d5ffd","url":"#b-94f3e6cb2e8f444b806d5ffd","relation":"figure-and-explanation"}]}]}