{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html","markdown":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.md","pages":[{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","unit":"检测分数很高，错误究竟在哪里？","unitId":"detection-errors-and-evidence","chapter":"II","number":1}],"aliases":[{"oldId":"b-73df7f5ed37f1932884caabf","targetId":"b-73df7f5ed37f1932884caabf","page":"reading-opening","operation":"move","beforeSHA256":"765ecc8f5e753bc61c0f41f7ffc4735144b20bd81fc7e76d0e664c0c463e01ee","afterSHA256":"765ecc8f5e753bc61c0f41f7ffc4735144b20bd81fc7e76d0e664c0c463e01ee","reason":"Same source body in authored semantic page"},{"oldId":"b-3ed891a01f1f97c483cb2346","targetId":"b-3ed891a01f1f97c483cb2346","page":"reading-opening","operation":"move","beforeSHA256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","afterSHA256":"6b45cf47a23e9eec2f15663b9a1624db57adbc4f2a6825dd6bc21c96bfb93aa4","reason":"Same source body in authored semantic page"},{"oldId":"b-7acd33e14a2c06ce35104a9b","targetId":"b-7acd33e14a2c06ce35104a9b","page":"reading-opening","operation":"move","beforeSHA256":"2721b637cb6f94558f6713c359032adafb78e252bc13f52e76ce70874291e64b","afterSHA256":"2721b637cb6f94558f6713c359032adafb78e252bc13f52e76ce70874291e64b","reason":"Same source body in authored semantic page"},{"oldId":"b-8df8a2dc36f950df2d73a279","targetId":"b-8df8a2dc36f950df2d73a279","page":"reading-opening","operation":"move","beforeSHA256":"7f64f21b960df2c59e03e9c18028330d1e9a641549b00585d29c5ca6e0f9b5b3","afterSHA256":"7f64f21b960df2c59e03e9c18028330d1e9a641549b00585d29c5ca6e0f9b5b3","reason":"Same source body in authored semantic page"},{"oldId":"b-c022901dea0f83dcefcd255e","targetId":"b-c022901dea0f83dcefcd255e","page":"reading-opening","operation":"rewrite","beforeSHA256":"0ca8335cec01613a77310772ede6cd2201e74b27c01699f9c6c5125faa8c5786","afterSHA256":"ea3f2b25b776460f4e0e045b665251387ea0450a36a4ff6475b9f0188e257153","reason":"Same source identity with figure semantics or reading context enriched"},{"oldId":"b-5007227ace1971097daa89c9","targetId":"b-5007227ace1971097daa89c9","page":"reading-opening","operation":"move","beforeSHA256":"8205b30be23a99f3c1540701846c79ab37442feeb48cf7193d02963131b7f9ab","afterSHA256":"8205b30be23a99f3c1540701846c79ab37442feeb48cf7193d02963131b7f9ab","reason":"Same source body in authored semantic page"},{"oldId":"b-f7b61e342c9ad6c60c60707c","targetId":"b-f7b61e342c9ad6c60c60707c","page":"reading-opening","operation":"move","beforeSHA256":"1caaa88a4dc7420a22d2566dc2b1a3e78a78af9d5f019befb84d7a50e0d3a296","afterSHA256":"1caaa88a4dc7420a22d2566dc2b1a3e78a78af9d5f019befb84d7a50e0d3a296","reason":"Same source body in authored semantic page"},{"oldId":"b-b8a54d7912b7b4d112e0987b","targetId":"b-b8a54d7912b7b4d112e0987b","page":"reading-opening","operation":"move","beforeSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","afterSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-73df7f5ed37f1932884caabf","type":"h1","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-73df7f5ed37f1932884caabf","content":"# 从 RGB 到框：当前 LAB10 已经具备完整闭环","sha256":"765ecc8f5e753bc61c0f41f7ffc4735144b20bd81fc7e76d0e664c0c463e01ee","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-3ed891a01f1f97c483cb2346","prerequisites":[{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-3ed891a01f1f97c483cb2346","type":"p","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-3ed891a01f1f97c483cb2346","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":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-73df7f5ed37f1932884caabf","nextBlock":"b-7acd33e14a2c06ce35104a9b","prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"}],"relatedBlocks":[]},{"id":"b-7acd33e14a2c06ce35104a9b","type":"p","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-7acd33e14a2c06ce35104a9b","content":"当前 LAB10 的主实验输入是真正的 48×48 RGB 合成图，目标是零至三个对象的类别与归一化框。卷积骨干生成 12×12 网格，每格两个参考框，共 288 个；两个头输出 `[B,288,3]` 类别 logits 与 `[B,288,4]` 偏移。图像进入 `model(images)`，坐标只参与监督与评价。","sha256":"2721b637cb6f94558f6713c359032adafb78e252bc13f52e76ce70874291e64b","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-3ed891a01f1f97c483cb2346","nextBlock":"b-8df8a2dc36f950df2d73a279","prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"},{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-8df8a2dc36f950df2d73a279","type":"p","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-8df8a2dc36f950df2d73a279","content":"保留的 2026-09-22 本地完整运行用 768 张训练图、96 张验证图、192 张保留图，训练 28 epoch。它以 TinySSD 的锚框检测闭环为起点，采用单特征层、两种参考尺度，不能称为 D2L 全部多尺度结构的复现。","sha256":"7f64f21b960df2c59e03e9c18028330d1e9a641549b00585d29c5ca6e0f9b5b3","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-7acd33e14a2c06ce35104a9b","nextBlock":"b-c022901dea0f83dcefcd255e","prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"},{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-c022901dea0f83dcefcd255e","type":"table","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-c022901dea0f83dcefcd255e","content":"| 条件 | AP50 | AP75 |\n| --- | --- | --- |\n| 随机初始化 | .00746 | .00003 |\n| 训练后原尺度 | .97303 | .70367 |\n| 同权重，仅对象尺寸减半 | .61490 | .17987 |","sha256":"ea3f2b25b776460f4e0e045b665251387ea0450a36a4ff6475b9f0188e257153","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"table":{"headers":["条件","AP50","AP75"],"rows":[["条件","AP50","AP75"],["随机初始化",".00746",".00003"],["训练后原尺度",".97303",".70367"],["同权重，仅对象尺寸减半",".61490",".17987"]]},"previousBlock":"b-8df8a2dc36f950df2d73a279","nextBlock":"b-5007227ace1971097daa89c9","prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"},{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5007227ace1971097daa89c9","type":"p","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-5007227ace1971097daa89c9","content":"这些是已有运行记录，本轮内容整理没有重跑实验。原分布明显改善，且存在多对象与空图，支持“模型学习到了图像到对象位置的映射”。更严格位置门槛与小尺度条件仍显著下降，帮助定位下一步研究问题。红色与矩形、蓝色与椭圆相关，因此不能据此宣称自然图像泛化。","sha256":"8205b30be23a99f3c1540701846c79ab37442feeb48cf7193d02963131b7f9ab","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-c022901dea0f83dcefcd255e","nextBlock":"b-f7b61e342c9ad6c60c60707c","prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"},{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-f7b61e342c9ad6c60c60707c","type":"p","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-f7b61e342c9ad6c60c60707c","content":"读者可先看同一行的 RGB、真值框、预测框和失败图，再看匹配与损失，最后按需运行。原四维逻辑回归保留在热身位置，它说明分类更新，不承担检测证据。","sha256":"1caaa88a4dc7420a22d2566dc2b1a3e78a78af9d5f019befb84d7a50e0d3a296","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-5007227ace1971097daa89c9","nextBlock":"b-b8a54d7912b7b4d112e0987b","prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"},{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-b8a54d7912b7b4d112e0987b","type":"h2","scope":"detector-learning-evidence","url":"https://codingai-lec03.pages.dev/course/blocks/detector-learning-evidence.html#b-b8a54d7912b7b4d112e0987b","content":"## 继续阅读与实验","sha256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","chapter":{"id":"II","title":"区域、像素与实例"},"unit":{"id":"detection-errors-and-evidence","title":"检测分数很高，错误究竟在哪里？"},"page":{"id":"reading-opening","title":"从 RGB 到框：当前 LAB10 已经具备完整闭环","number":1},"sectionTitle":"从 RGB 到框：当前 LAB10 已经具备完整闭环","currentQuestion":"从 RGB 到框：当前 LAB10 已经具备完整闭环","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-f7b61e342c9ad6c60c60707c","nextBlock":null,"prerequisites":[{"id":"b-73df7f5ed37f1932884caabf","url":"#b-73df7f5ed37f1932884caabf","role":"section-title"},{"id":"b-3ed891a01f1f97c483cb2346","url":"#b-3ed891a01f1f97c483cb2346","role":"problem-context"}],"relatedBlocks":[]}]}