{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"选读：一阶段检测保留了哪些相同问题","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html","markdown":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.md","pages":[{"id":"reading-opening","title":"选读：一阶段检测保留了哪些相同问题","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-7d8fb7dad2623e1851a77ee8","targetId":"b-7d8fb7dad2623e1851a77ee8","page":"reading-opening","operation":"move","beforeSHA256":"657badb13bb27f054e29e6c011fea118c2f1982fb54f4a67617278d397c9cbf0","afterSHA256":"657badb13bb27f054e29e6c011fea118c2f1982fb54f4a67617278d397c9cbf0","reason":"Same source body in authored semantic page"},{"oldId":"b-1d1894fef6f081e28cc012ea","targetId":"b-1d1894fef6f081e28cc012ea","page":"reading-opening","operation":"move","beforeSHA256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","afterSHA256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","reason":"Same source body in authored semantic page"},{"oldId":"b-3ffde533ecb135ec8818debb","targetId":"b-3ffde533ecb135ec8818debb","page":"reading-opening","operation":"move","beforeSHA256":"c3fba1f619d9fa82c144218a2e5408f99396070490631920a72f7ead7b84d288","afterSHA256":"c3fba1f619d9fa82c144218a2e5408f99396070490631920a72f7ead7b84d288","reason":"Same source body in authored semantic page"},{"oldId":"b-a0b69e9476b04de6413f1f34","targetId":"b-a0b69e9476b04de6413f1f34","page":"reading-opening","operation":"move","beforeSHA256":"7d8fe2f5f510715c2b69ed15383f5797f594930db9819d67d883ae2d3c44722e","afterSHA256":"7d8fe2f5f510715c2b69ed15383f5797f594930db9819d67d883ae2d3c44722e","reason":"Same source body in authored semantic page"},{"oldId":"b-534c8ec5b388cda2d1f30c3a","targetId":"b-534c8ec5b388cda2d1f30c3a","page":"reading-opening","operation":"move","beforeSHA256":"159608d00faa18d3a91c3edd497cb1a181cb69965e0d2c41b1161b30140fa3ae","afterSHA256":"159608d00faa18d3a91c3edd497cb1a181cb69965e0d2c41b1161b30140fa3ae","reason":"Same source body in authored semantic page"},{"oldId":"b-f1441ed63690fb1129d87ad6","targetId":"b-f1441ed63690fb1129d87ad6","page":"reading-opening","operation":"move","beforeSHA256":"6b8c8150ee61580fb752c1fde1e79a70c46e4b0ffcd07319b820b1e4ce5c430e","afterSHA256":"6b8c8150ee61580fb752c1fde1e79a70c46e4b0ffcd07319b820b1e4ce5c430e","reason":"Same source body in authored semantic page"},{"oldId":"b-6972b0167adf6bbcb5e5c3a4","targetId":"b-6972b0167adf6bbcb5e5c3a4","page":"reading-opening","operation":"move","beforeSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","afterSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-7d8fb7dad2623e1851a77ee8","type":"h1","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-7d8fb7dad2623e1851a77ee8","content":"# 选读：一阶段检测保留了哪些相同问题","sha256":"657badb13bb27f054e29e6c011fea118c2f1982fb54f4a67617278d397c9cbf0","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-1d1894fef6f081e28cc012ea","prerequisites":[{"id":"b-1d1894fef6f081e28cc012ea","url":"#b-1d1894fef6f081e28cc012ea","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-1d1894fef6f081e28cc012ea","type":"p","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-1d1894fef6f081e28cc012ea","content":"按“参考框→RPN 候选→最终框”给图中的矩形分配身份，并把人工真值单独留在监督侧。RPN 让候选变为可学习预测。LAB10 采用共享原理的一阶段教学变体，不应把它说成原版 Faster R-CNN。","sha256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-7d8fb7dad2623e1851a77ee8","nextBlock":"b-3ffde533ecb135ec8818debb","prerequisites":[{"id":"b-7d8fb7dad2623e1851a77ee8","url":"#b-7d8fb7dad2623e1851a77ee8","role":"section-title"}],"relatedBlocks":[]},{"id":"b-3ffde533ecb135ec8818debb","type":"p","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-3ffde533ecb135ec8818debb","content":"一阶段检测不先为每个 proposal 再做一套区域分类，而是在一个或多个特征网格上直接预测类别与位置。这个改变简化了输出路径，却没有消除参考几何、目标匹配、正负不平衡、重复预测与定位评价。","sha256":"c3fba1f619d9fa82c144218a2e5408f99396070490631920a72f7ead7b84d288","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-1d1894fef6f081e28cc012ea","nextBlock":"b-a0b69e9476b04de6413f1f34","prerequisites":[{"id":"b-7d8fb7dad2623e1851a77ee8","url":"#b-7d8fb7dad2623e1851a77ee8","role":"section-title"},{"id":"b-1d1894fef6f081e28cc012ea","url":"#b-1d1894fef6f081e28cc012ea","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-a0b69e9476b04de6413f1f34","type":"p","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-a0b69e9476b04de6413f1f34","content":"若 16×16 网格每格有 4 个参考框，共 1024 项；每项预测含背景的 3 类 logits 与 4 个偏移，就有 7168 个输出数。这些数还是结构化候选，需按同一位置顺序与参考框配对，随后解码并整理。张量 reshape 顺序错误时，模型的偏移可能被贴到另一个 anchor 上。","sha256":"7d8fe2f5f510715c2b69ed15383f5797f594930db9819d67d883ae2d3c44722e","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-3ffde533ecb135ec8818debb","nextBlock":"b-534c8ec5b388cda2d1f30c3a","prerequisites":[{"id":"b-7d8fb7dad2623e1851a77ee8","url":"#b-7d8fb7dad2623e1851a77ee8","role":"section-title"},{"id":"b-1d1894fef6f081e28cc012ea","url":"#b-1d1894fef6f081e28cc012ea","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-534c8ec5b388cda2d1f30c3a","type":"p","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-534c8ec5b388cda2d1f30c3a","content":"D2L TinySSD 在多尺度特征上构造参考框与预测。当前 LAB10 为保持 CPU 可运行，用 12×12 单层网格、每格两个尺寸，共 288 项；它保留图像输入到框输出的学习链，但不等于多尺度模型的全部能力。","sha256":"159608d00faa18d3a91c3edd497cb1a181cb69965e0d2c41b1161b30140fa3ae","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-a0b69e9476b04de6413f1f34","nextBlock":"b-f1441ed63690fb1129d87ad6","prerequisites":[{"id":"b-7d8fb7dad2623e1851a77ee8","url":"#b-7d8fb7dad2623e1851a77ee8","role":"section-title"},{"id":"b-1d1894fef6f081e28cc012ea","url":"#b-1d1894fef6f081e28cc012ea","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-f1441ed63690fb1129d87ad6","type":"p","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-f1441ed63690fb1129d87ad6","content":"本条作为从主讲移出的深入阅读：理解它有助于读实验代码，不要求课堂再开 SSD 全章。FPN 关注怎样形成不同尺度的表示，SSD 关注怎样在特征上预测；两者都出现多尺度，不意味着同一机制。","sha256":"6b8c8150ee61580fb752c1fde1e79a70c46e4b0ffcd07319b820b1e4ce5c430e","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-534c8ec5b388cda2d1f30c3a","nextBlock":"b-6972b0167adf6bbcb5e5c3a4","prerequisites":[{"id":"b-7d8fb7dad2623e1851a77ee8","url":"#b-7d8fb7dad2623e1851a77ee8","role":"section-title"},{"id":"b-1d1894fef6f081e28cc012ea","url":"#b-1d1894fef6f081e28cc012ea","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-6972b0167adf6bbcb5e5c3a4","type":"h2","scope":"dense-detector-ssd-extension","url":"https://codingai-lec03.pages.dev/course/blocks/dense-detector-ssd-extension.html#b-6972b0167adf6bbcb5e5c3a4","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":"选读：一阶段检测保留了哪些相同问题","number":1},"sectionTitle":"选读：一阶段检测保留了哪些相同问题","currentQuestion":"选读：一阶段检测保留了哪些相同问题","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-f1441ed63690fb1129d87ad6","nextBlock":null,"prerequisites":[{"id":"b-7d8fb7dad2623e1851a77ee8","url":"#b-7d8fb7dad2623e1851a77ee8","role":"section-title"},{"id":"b-1d1894fef6f081e28cc012ea","url":"#b-1d1894fef6f081e28cc012ea","role":"problem-context"}],"relatedBlocks":[]}]}