{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"FCN 从密集像素预测出发","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.md","pages":[{"id":"reading-opening","title":"FCN 从密集像素预测出发","unit":"FCN 从密集像素预测出发","unitId":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","chapter":"course","number":1}],"aliases":[{"oldId":"b-72014e71bb3d753d49640f0e","targetId":"b-72014e71bb3d753d49640f0e","page":"reading-opening","operation":"move","beforeSHA256":"923e066494e483e0d0fdd514ec12b59fe7659670980fdf4639d15a488eccad0e","afterSHA256":"923e066494e483e0d0fdd514ec12b59fe7659670980fdf4639d15a488eccad0e","reason":"Same source body in authored semantic page"},{"oldId":"b-6b171a73f3d62825778f7186","targetId":"b-6b171a73f3d62825778f7186","page":"reading-opening","operation":"move","beforeSHA256":"27882f25025d68e7425383bc2f264dbf30a8977ceb25868c2bdaf9c19f9c46a1","afterSHA256":"27882f25025d68e7425383bc2f264dbf30a8977ceb25868c2bdaf9c19f9c46a1","reason":"Same source body in authored semantic page"},{"oldId":"b-6bba32b1448169bee88c5089","targetId":"b-6bba32b1448169bee88c5089","page":"reading-opening","operation":"move","beforeSHA256":"42cba31deccba2a350b9b23bd0ca1dcaa4c4c5cb9d35c773e1fcb89124788ccd","afterSHA256":"42cba31deccba2a350b9b23bd0ca1dcaa4c4c5cb9d35c773e1fcb89124788ccd","reason":"Same source body in authored semantic page"},{"oldId":"b-e2b9f2caa4c593b3939b14c8","targetId":"b-e2b9f2caa4c593b3939b14c8","page":"reading-opening","operation":"move","beforeSHA256":"e0584a8ed4ddfb5b9c9abb132012919eb506480c39698c3ddd67fccd73d0a12e","afterSHA256":"e0584a8ed4ddfb5b9c9abb132012919eb506480c39698c3ddd67fccd73d0a12e","reason":"Same source body in authored semantic page"},{"oldId":"b-27d55ab24721b82f551713e8","targetId":"b-27d55ab24721b82f551713e8","page":"reading-opening","operation":"move","beforeSHA256":"c4495cfa2bc9337007b9041d1f1c21ec4758b4e475f79bd830a80212b304b991","afterSHA256":"c4495cfa2bc9337007b9041d1f1c21ec4758b4e475f79bd830a80212b304b991","reason":"Same source body in authored semantic page"},{"oldId":"b-b07f94220b5c44ac80272c11","targetId":"b-b07f94220b5c44ac80272c11","page":"reading-opening","operation":"move","beforeSHA256":"f7a35b191037a3d4ce1a0380354829523a6771e8f0f817b75d9680684ce80630","afterSHA256":"f7a35b191037a3d4ce1a0380354829523a6771e8f0f817b75d9680684ce80630","reason":"Same source body in authored semantic page"},{"oldId":"b-eb554b069f70674f1b0d2cf4","targetId":"b-eb554b069f70674f1b0d2cf4","page":"reading-opening","operation":"move","beforeSHA256":"577437aff921fbc4647e84807f5595582e274dea1f87371f95b51589c0894d0c","afterSHA256":"577437aff921fbc4647e84807f5595582e274dea1f87371f95b51589c0894d0c","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-72014e71bb3d753d49640f0e","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-72014e71bb3d753d49640f0e","content":"[CLS 02 · 从类别到区域与掩码](/course/notebooks/f9ab632ba16b4f2084669791cba6ba85.html)","sha256":"923e066494e483e0d0fdd514ec12b59fe7659670980fdf4639d15a488eccad0e","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-6b171a73f3d62825778f7186","prerequisites":[],"relatedBlocks":[]},{"id":"b-6b171a73f3d62825778f7186","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-6b171a73f3d62825778f7186","content":"# FCN 从密集像素预测出发","sha256":"27882f25025d68e7425383bc2f264dbf30a8977ceb25868c2bdaf9c19f9c46a1","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-72014e71bb3d753d49640f0e","nextBlock":"b-6bba32b1448169bee88c5089","prerequisites":[{"id":"b-72014e71bb3d753d49640f0e","url":"#b-72014e71bb3d753d49640f0e","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-6bba32b1448169bee88c5089","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-6bba32b1448169bee88c5089","content":"主线阅读与优先实验","sha256":"42cba31deccba2a350b9b23bd0ca1dcaa4c4c5cb9d35c773e1fcb89124788ccd","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6b171a73f3d62825778f7186","nextBlock":"b-e2b9f2caa4c593b3939b14c8","prerequisites":[{"id":"b-6b171a73f3d62825778f7186","url":"#b-6b171a73f3d62825778f7186","role":"section-title"},{"id":"b-72014e71bb3d753d49640f0e","url":"#b-72014e71bb3d753d49640f0e","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-e2b9f2caa4c593b3939b14c8","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-e2b9f2caa4c593b3939b14c8","content":"同图四输出、共享计算、FCN、RoIAlign、信息碰撞与SAM接口","sha256":"e0584a8ed4ddfb5b9c9abb132012919eb506480c39698c3ddd67fccd73d0a12e","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6bba32b1448169bee88c5089","nextBlock":"b-27d55ab24721b82f551713e8","prerequisites":[{"id":"b-6b171a73f3d62825778f7186","url":"#b-6b171a73f3d62825778f7186","role":"section-title"},{"id":"b-72014e71bb3d753d49640f0e","url":"#b-72014e71bb3d753d49640f0e","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-27d55ab24721b82f551713e8","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-27d55ab24721b82f551713e8","content":"FCN 从密集像素预测出发。网络在整个图像上产生类别分数图，再恢复到与输入相对应的输出尺度。深层特征容易汇聚较大范围的语义信息，但空间网格较粗；较浅层保留更细的位置线索。跳跃连接把这些层次的线索结合起来。[FCN](https://arxiv.org/abs/1411.4038)","sha256":"c4495cfa2bc9337007b9041d1f1c21ec4758b4e475f79bd830a80212b304b991","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-e2b9f2caa4c593b3939b14c8","nextBlock":"b-b07f94220b5c44ac80272c11","prerequisites":[{"id":"b-6b171a73f3d62825778f7186","url":"#b-6b171a73f3d62825778f7186","role":"section-title"},{"id":"b-72014e71bb3d753d49640f0e","url":"#b-72014e71bb3d753d49640f0e","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-b07f94220b5c44ac80272c11","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-b07f94220b5c44ac80272c11","content":"对于两只杯子，语义分割可以把它们都标为“杯子”。然而一块连通的杯子像素既可能是一只宽杯，也可能是两只相接的杯子。类别值本身不规定实例数量。","sha256":"f7a35b191037a3d4ce1a0380354829523a6771e8f0f817b75d9680684ce80630","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-27d55ab24721b82f551713e8","nextBlock":"b-eb554b069f70674f1b0d2cf4","prerequisites":[{"id":"b-6b171a73f3d62825778f7186","url":"#b-6b171a73f3d62825778f7186","role":"section-title"},{"id":"b-72014e71bb3d753d49640f0e","url":"#b-72014e71bb3d753d49640f0e","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-eb554b069f70674f1b0d2cf4","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/e07a99e024e54a96b95494f93248d35f.html#b-eb554b069f70674f1b0d2cf4","content":"因此有两条问题线：**区域路线组织“哪些对象”与框；密集路线组织“哪些像素”与类别。** 它们可以汇合，但不是沿一条箭头依次淘汰。","sha256":"577437aff921fbc4647e84807f5595582e274dea1f87371f95b51589c0894d0c","chapter":{"id":"course","title":"FCN 从密集像素预测出发"},"unit":{"id":"notebook-blocks/e07a99e024e54a96b95494f93248d35f","title":"FCN 从密集像素预测出发"},"page":{"id":"reading-opening","title":"FCN 从密集像素预测出发","number":1},"sectionTitle":"FCN 从密集像素预测出发","currentQuestion":"FCN 从密集像素预测出发","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-b07f94220b5c44ac80272c11","nextBlock":null,"prerequisites":[{"id":"b-6b171a73f3d62825778f7186","url":"#b-6b171a73f3d62825778f7186","role":"section-title"},{"id":"b-72014e71bb3d753d49640f0e","url":"#b-72014e71bb3d753d49640f0e","role":"problem-context"}],"relatedBlocks":[]}]}