{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"为什么同时训练 1×1 读出和 FCN","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.md","pages":[{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","unit":"为什么同时训练 1×1 读出和 FCN","unitId":"notebook-blocks/a83492c549764a6aad90a23cfb322167","chapter":"course","number":1}],"aliases":[{"oldId":"b-6901f1f6fea6a896aa0c2b14","targetId":"b-6901f1f6fea6a896aa0c2b14","page":"reading-opening","operation":"move","beforeSHA256":"f6dfd09a50caab375f02e0abcc3ff7a454ea579ed8eef0c35b95c95b0619c304","afterSHA256":"f6dfd09a50caab375f02e0abcc3ff7a454ea579ed8eef0c35b95c95b0619c304","reason":"Same source body in authored semantic page"},{"oldId":"b-5ea420afa64f77da15a9e0f2","targetId":"b-5ea420afa64f77da15a9e0f2","page":"reading-opening","operation":"move","beforeSHA256":"eec753d1536650aecda69557c4ab68c04bfd2918c43fd9064011e1efbeb95fdb","afterSHA256":"eec753d1536650aecda69557c4ab68c04bfd2918c43fd9064011e1efbeb95fdb","reason":"Same source body in authored semantic page"},{"oldId":"b-84f0a02e976d7282f5eb1547","targetId":"b-84f0a02e976d7282f5eb1547","page":"reading-opening","operation":"move","beforeSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","afterSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","reason":"Same source body in authored semantic page"},{"oldId":"b-a8aef558e720ea51bbc83c9e","targetId":"b-a8aef558e720ea51bbc83c9e","page":"reading-opening","operation":"move","beforeSHA256":"6f0b115c3c32032428a4c0d5ec4791e16d6b63aa3d643457bed758bf6af2b765","afterSHA256":"6f0b115c3c32032428a4c0d5ec4791e16d6b63aa3d643457bed758bf6af2b765","reason":"Same source body in authored semantic page"},{"oldId":"b-802c4822584617fac61057e0","targetId":"b-802c4822584617fac61057e0","page":"reading-opening","operation":"move","beforeSHA256":"d47e56d245656ae887c7eb760e578713ac36b028b1761e7a1cdc24c593cc57b1","afterSHA256":"d47e56d245656ae887c7eb760e578713ac36b028b1761e7a1cdc24c593cc57b1","reason":"Same source body in authored semantic page"},{"oldId":"b-e37db2fd9094367ad10ae42f","targetId":"b-e37db2fd9094367ad10ae42f","page":"reading-opening","operation":"move","beforeSHA256":"7d5a1add654bd29f2bf8645c182e7b93815814ebedcfc5d3f9c65a7686918d4c","afterSHA256":"7d5a1add654bd29f2bf8645c182e7b93815814ebedcfc5d3f9c65a7686918d4c","reason":"Same source body in authored semantic page"},{"oldId":"b-5365534347e3a90b1fd4777c","targetId":"b-5365534347e3a90b1fd4777c","page":"reading-opening","operation":"move","beforeSHA256":"2c69e23b066df28a6a6aba3050f0068b12c0749c5772e39e19e4960b4ba27ba0","afterSHA256":"2c69e23b066df28a6a6aba3050f0068b12c0749c5772e39e19e4960b4ba27ba0","reason":"Same source body in authored semantic page"},{"oldId":"b-5e97cfe0b08a57170b887aa3","targetId":"b-5e97cfe0b08a57170b887aa3","page":"reading-opening","operation":"move","beforeSHA256":"fe711ade7476462f1282523a3800320717ac116f9f5f229c3faab01f52220735","afterSHA256":"fe711ade7476462f1282523a3800320717ac116f9f5f229c3faab01f52220735","reason":"Same source body in authored semantic page"},{"oldId":"block-a83492c549764a6aad90a23cfb322167--为什么同时训练-11-读出和-fcn","targetId":"b-802c4822584617fac61057e0","operation":"retain-source-identity"}],"blocks":[{"id":"b-6901f1f6fea6a896aa0c2b14","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-6901f1f6fea6a896aa0c2b14","content":"[LAB 11 · 从 RGB 到像素：读出、FCN 与边界](/course/notebooks/9e268706471944e3a59e6f568f9ce9a3.html)","sha256":"f6dfd09a50caab375f02e0abcc3ff7a454ea579ed8eef0c35b95c95b0619c304","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-5ea420afa64f77da15a9e0f2","prerequisites":[],"relatedBlocks":[]},{"id":"b-5ea420afa64f77da15a9e0f2","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-5ea420afa64f77da15a9e0f2","content":"# 为什么同时训练 1×1 读出和 FCN","sha256":"eec753d1536650aecda69557c4ab68c04bfd2918c43fd9064011e1efbeb95fdb","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6901f1f6fea6a896aa0c2b14","nextBlock":"b-84f0a02e976d7282f5eb1547","prerequisites":[{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-84f0a02e976d7282f5eb1547","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-84f0a02e976d7282f5eb1547","content":"按问题继续","sha256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-5ea420afa64f77da15a9e0f2","nextBlock":"b-a8aef558e720ea51bbc83c9e","prerequisites":[{"id":"b-5ea420afa64f77da15a9e0f2","url":"#b-5ea420afa64f77da15a9e0f2","role":"section-title"},{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-a8aef558e720ea51bbc83c9e","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-a8aef558e720ea51bbc83c9e","content":"前景/背景/mIoU口径、RGB到mask训练、FCN与1×1读出及边界","sha256":"6f0b115c3c32032428a4c0d5ec4791e16d6b63aa3d643457bed758bf6af2b765","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-84f0a02e976d7282f5eb1547","nextBlock":"b-802c4822584617fac61057e0","prerequisites":[{"id":"b-5ea420afa64f77da15a9e0f2","url":"#b-5ea420afa64f77da15a9e0f2","role":"section-title"},{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-802c4822584617fac61057e0","type":"h2","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-802c4822584617fac61057e0","content":"## 为什么同时训练 1×1 读出和 FCN","sha256":"d47e56d245656ae887c7eb760e578713ac36b028b1761e7a1cdc24c593cc57b1","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-a8aef558e720ea51bbc83c9e","nextBlock":"b-e37db2fd9094367ad10ae42f","prerequisites":[{"id":"b-5ea420afa64f77da15a9e0f2","url":"#b-5ea420afa64f77da15a9e0f2","role":"section-title"},{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-e37db2fd9094367ad10ae42f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-e37db2fd9094367ad10ae42f","content":"逐像素读出是 `Conv2d(3,3,1)`，只有 12 个参数：每个像素用共享线性变换和 softmax 分类，没有邻域、形状或对象关系。它是原正文曾承诺而旧代码未实现的可学习逻辑读出。","sha256":"7d5a1add654bd29f2bf8645c182e7b93815814ebedcfc5d3f9c65a7686918d4c","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-802c4822584617fac61057e0","nextBlock":"b-5365534347e3a90b1fd4777c","prerequisites":[{"id":"b-802c4822584617fac61057e0","url":"#b-802c4822584617fac61057e0","role":"section-title"},{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5365534347e3a90b1fd4777c","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-5365534347e3a90b1fd4777c","content":"FCN 有 20,590 个参数。细分支在 48×48 上提取边缘；粗分支两次 stride=2 得到 12×12 特征，再用 1×1 卷积产生三通道 logits。粗 logits 双线性恢复到 48×48，与细分支 logits 相加。细分支保留空间细节，粗分支提供较大邻域；两者通过同一个像素交叉熵一起更新。这个结构借鉴 FCN 的粗细融合思想，并非 D2L 预训练 ResNet-18/转置卷积实现的逐行复制。","sha256":"2c69e23b066df28a6a6aba3050f0068b12c0749c5772e39e19e4960b4ba27ba0","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-e37db2fd9094367ad10ae42f","nextBlock":"b-5e97cfe0b08a57170b887aa3","prerequisites":[{"id":"b-802c4822584617fac61057e0","url":"#b-802c4822584617fac61057e0","role":"section-title"},{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5e97cfe0b08a57170b887aa3","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/a83492c549764a6aad90a23cfb322167.html#b-5e97cfe0b08a57170b887aa3","content":"损失 $-\\sum_{u\\in valid}\\log softmax(z_u)_{y_u}/|valid|$ 对所有有效像素平均。没有任何一项要求不同物体的实例 ID，因此相邻的两个蓝椭圆都应是类别 2；但椭圆间属于背景的窄缝仍必须保留。","sha256":"fe711ade7476462f1282523a3800320717ac116f9f5f229c3faab01f52220735","chapter":{"id":"course","title":"为什么同时训练 1×1 读出和 FCN"},"unit":{"id":"notebook-blocks/a83492c549764a6aad90a23cfb322167","title":"为什么同时训练 1×1 读出和 FCN"},"page":{"id":"reading-opening","title":"为什么同时训练 1×1 读出和 FCN","number":1},"sectionTitle":"为什么同时训练 1×1 读出和 FCN","currentQuestion":"为什么同时训练 1×1 读出和 FCN","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"-\\sum_{u\\in valid}\\log softmax(z_u)_{y_u}/|valid|","definitionContext":"为什么同时训练 1×1 读出和 FCN"}],"previousBlock":"b-5365534347e3a90b1fd4777c","nextBlock":null,"prerequisites":[{"id":"b-802c4822584617fac61057e0","url":"#b-802c4822584617fac61057e0","role":"section-title"},{"id":"b-6901f1f6fea6a896aa0c2b14","url":"#b-6901f1f6fea6a896aa0c2b14","role":"problem-context"}],"relatedBlocks":[]}]}