{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"为什么要有正例、背景和忽略项","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html","markdown":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.md","pages":[{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","unit":"类别与位置怎样成为可学习的区域目标？","unitId":"learnable-box-prediction","chapter":"II","number":1}],"aliases":[{"oldId":"b-690e6ba1a8b3f5d839ee8e45","targetId":"b-690e6ba1a8b3f5d839ee8e45","page":"reading-opening","operation":"move","beforeSHA256":"05429cffc05d436c75ed7ed1d55275494e7d528ee98811ad8a77627c62e7fb90","afterSHA256":"05429cffc05d436c75ed7ed1d55275494e7d528ee98811ad8a77627c62e7fb90","reason":"Same source body in authored semantic page"},{"oldId":"b-7bb20faf8487a00d64241378","targetId":"b-7bb20faf8487a00d64241378","page":"reading-opening","operation":"move","beforeSHA256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","afterSHA256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","reason":"Same source body in authored semantic page"},{"oldId":"b-11c037765baf14d9ca04f15c","targetId":"b-11c037765baf14d9ca04f15c","page":"reading-opening","operation":"move","beforeSHA256":"290dba49c527e68e2aba4a2ee0b4d28e626d8e489edab4605ea5a235e8ea7fb2","afterSHA256":"290dba49c527e68e2aba4a2ee0b4d28e626d8e489edab4605ea5a235e8ea7fb2","reason":"Same source body in authored semantic page"},{"oldId":"b-6cf31747ac5f8b44780af44e","targetId":"b-6cf31747ac5f8b44780af44e","page":"reading-opening","operation":"move","beforeSHA256":"9950be4406a476d8d766763df53ea513bbe7eed02f34aff4f837c30a3d045b5c","afterSHA256":"9950be4406a476d8d766763df53ea513bbe7eed02f34aff4f837c30a3d045b5c","reason":"Same source body in authored semantic page"},{"oldId":"b-5b3740a4c3c40bb1b6472270","targetId":"b-5b3740a4c3c40bb1b6472270","page":"reading-opening","operation":"move","beforeSHA256":"157cb5eec4f8576589386132685fedffe3cc7d7ba664168060367d6e5a58232b","afterSHA256":"157cb5eec4f8576589386132685fedffe3cc7d7ba664168060367d6e5a58232b","reason":"Same source body in authored semantic page"},{"oldId":"b-028ae8dec5ed0ee5e012ee0e","targetId":"b-028ae8dec5ed0ee5e012ee0e","page":"reading-opening","operation":"move","beforeSHA256":"767e9cdba079c3d5360d01ca6637ff47b625f805ea988b0a945b3600317d283c","afterSHA256":"767e9cdba079c3d5360d01ca6637ff47b625f805ea988b0a945b3600317d283c","reason":"Same source body in authored semantic page"},{"oldId":"b-3190e8c1e2d671694a1c3f29","targetId":"b-3190e8c1e2d671694a1c3f29","page":"reading-opening","operation":"move","beforeSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","afterSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-690e6ba1a8b3f5d839ee8e45","type":"h1","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-690e6ba1a8b3f5d839ee8e45","content":"# 为什么要有正例、背景和忽略项","sha256":"05429cffc05d436c75ed7ed1d55275494e7d528ee98811ad8a77627c62e7fb90","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-7bb20faf8487a00d64241378","prerequisites":[{"id":"b-7bb20faf8487a00d64241378","url":"#b-7bb20faf8487a00d64241378","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-7bb20faf8487a00d64241378","type":"p","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-7bb20faf8487a00d64241378","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":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-690e6ba1a8b3f5d839ee8e45","nextBlock":"b-11c037765baf14d9ca04f15c","prerequisites":[{"id":"b-690e6ba1a8b3f5d839ee8e45","url":"#b-690e6ba1a8b3f5d839ee8e45","role":"section-title"}],"relatedBlocks":[]},{"id":"b-11c037765baf14d9ca04f15c","type":"p","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-11c037765baf14d9ca04f15c","content":"给定参考框与真值的 IoU 矩阵，匹配规则把参考框变成训练样本。例如课堂教学阈值若设正例至少 0.7、背景小于 0.3，则 IoU 为 0.8、0.4、0.1 的三个框分别是正例、忽略、背景。只有正例有对应的框回归目标。忽略的用途是暂不把模糊覆盖强行当正或负，而不是悄悄改成背景。","sha256":"290dba49c527e68e2aba4a2ee0b4d28e626d8e489edab4605ea5a235e8ea7fb2","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-7bb20faf8487a00d64241378","nextBlock":"b-6cf31747ac5f8b44780af44e","prerequisites":[{"id":"b-690e6ba1a8b3f5d839ee8e45","url":"#b-690e6ba1a8b3f5d839ee8e45","role":"section-title"},{"id":"b-7bb20faf8487a00d64241378","url":"#b-7bb20faf8487a00d64241378","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-6cf31747ac5f8b44780af44e","type":"p","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-6cf31747ac5f8b44780af44e","content":"真实 LAB10 使用另一组公开写明的阈值：正例至少 0.45，背景低于 0.25，中间忽略；再用逐次最大 IoU 为每个真值保留一个不同 anchor，避免某个对象完全收不到位置监督。强制匹配服务于学习，评价时不能为了“每个对象都匹配一次”照搬这条规则。","sha256":"9950be4406a476d8d766763df53ea513bbe7eed02f34aff4f837c30a3d045b5c","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-11c037765baf14d9ca04f15c","nextBlock":"b-5b3740a4c3c40bb1b6472270","prerequisites":[{"id":"b-690e6ba1a8b3f5d839ee8e45","url":"#b-690e6ba1a8b3f5d839ee8e45","role":"section-title"},{"id":"b-7bb20faf8487a00d64241378","url":"#b-7bb20faf8487a00d64241378","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5b3740a4c3c40bb1b6472270","type":"p","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-5b3740a4c3c40bb1b6472270","content":"空图同样是有效训练输入。所有参考框都应学背景，回归项没有正例便归零。如果只训练总有物体的图片，模型可能从未被要求学会空输出。","sha256":"157cb5eec4f8576589386132685fedffe3cc7d7ba664168060367d6e5a58232b","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6cf31747ac5f8b44780af44e","nextBlock":"b-028ae8dec5ed0ee5e012ee0e","prerequisites":[{"id":"b-690e6ba1a8b3f5d839ee8e45","url":"#b-690e6ba1a8b3f5d839ee8e45","role":"section-title"},{"id":"b-7bb20faf8487a00d64241378","url":"#b-7bb20faf8487a00d64241378","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-028ae8dec5ed0ee5e012ee0e","type":"p","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-028ae8dec5ed0ee5e012ee0e","content":"负例远多于正例时，本实验选难负例参与分类损失，最多正例数的三倍、每图至少十二个。阅读代码时应看 loss 的掩码与归一化，而不只看 $L=L_{cls}+\\lambda L_{box}$ 这一行总公式。","sha256":"767e9cdba079c3d5360d01ca6637ff47b625f805ea988b0a945b3600317d283c","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"L=L_{cls}+\\lambda L_{box}","definitionContext":"为什么要有正例、背景和忽略项"}],"previousBlock":"b-5b3740a4c3c40bb1b6472270","nextBlock":"b-3190e8c1e2d671694a1c3f29","prerequisites":[{"id":"b-690e6ba1a8b3f5d839ee8e45","url":"#b-690e6ba1a8b3f5d839ee8e45","role":"section-title"},{"id":"b-7bb20faf8487a00d64241378","url":"#b-7bb20faf8487a00d64241378","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-3190e8c1e2d671694a1c3f29","type":"h2","scope":"anchor-training-assignment","url":"https://codingai-lec03.pages.dev/course/blocks/anchor-training-assignment.html#b-3190e8c1e2d671694a1c3f29","content":"## 继续阅读与实验","sha256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"learnable-box-prediction","title":"类别与位置怎样成为可学习的区域目标？"},"page":{"id":"reading-opening","title":"为什么要有正例、背景和忽略项","number":1},"sectionTitle":"为什么要有正例、背景和忽略项","currentQuestion":"为什么要有正例、背景和忽略项","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-028ae8dec5ed0ee5e012ee0e","nextBlock":null,"prerequisites":[{"id":"b-690e6ba1a8b3f5d839ee8e45","url":"#b-690e6ba1a8b3f5d839ee8e45","role":"section-title"},{"id":"b-7bb20faf8487a00d64241378","url":"#b-7bb20faf8487a00d64241378","role":"problem-context"}],"relatedBlocks":[]}]}