{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"一张类别图怎样成为可训练的输出","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.md","pages":[{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","unit":"一张类别图怎样成为可训练的输出","unitId":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","chapter":"course","number":1}],"aliases":[{"oldId":"b-f3dc1ec29e18aec577e14dda","targetId":"b-f3dc1ec29e18aec577e14dda","page":"reading-opening","operation":"move","beforeSHA256":"83aaf240c580bb5e1a3c458920748ed89991c349313e733e2c8bb2d5ae6ac25d","afterSHA256":"83aaf240c580bb5e1a3c458920748ed89991c349313e733e2c8bb2d5ae6ac25d","reason":"Same source body in authored semantic page"},{"oldId":"b-8a65403a37d4bbd2d691096b","targetId":"b-8a65403a37d4bbd2d691096b","page":"reading-opening","operation":"move","beforeSHA256":"6afaf53c5c378b64998573e505c11a4a5921e1c2b4ec5292171acc7ce6507a3c","afterSHA256":"6afaf53c5c378b64998573e505c11a4a5921e1c2b4ec5292171acc7ce6507a3c","reason":"Same source body in authored semantic page"},{"oldId":"b-0863ea013ba4265d69619fab","targetId":"b-0863ea013ba4265d69619fab","page":"reading-opening","operation":"move","beforeSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","afterSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","reason":"Same source body in authored semantic page"},{"oldId":"b-c561639ba6edfdbb0eb7aee8","targetId":"b-c561639ba6edfdbb0eb7aee8","page":"reading-opening","operation":"move","beforeSHA256":"70eec56f79b131d4695a1c420bc695a02ef3863727481994de8f7ed4bd673dcc","afterSHA256":"70eec56f79b131d4695a1c420bc695a02ef3863727481994de8f7ed4bd673dcc","reason":"Same source body in authored semantic page"},{"oldId":"b-802b09f0ef7cd549038d90a7","targetId":"b-802b09f0ef7cd549038d90a7","page":"reading-opening","operation":"move","beforeSHA256":"30ab9f304366670818536cf73c98c66aa814e8e31ecdbffcf19a4d6e7e7f22de","afterSHA256":"30ab9f304366670818536cf73c98c66aa814e8e31ecdbffcf19a4d6e7e7f22de","reason":"Same source body in authored semantic page"},{"oldId":"b-29e04a304ea18f0cb8f0f6bd","targetId":"b-29e04a304ea18f0cb8f0f6bd","page":"reading-opening","operation":"move","beforeSHA256":"72798cbfc37e71e8b9dbbcfe3cfb8a7b0fc2d74531b251cdcb3e0c9989b9c37d","afterSHA256":"72798cbfc37e71e8b9dbbcfe3cfb8a7b0fc2d74531b251cdcb3e0c9989b9c37d","reason":"Same source body in authored semantic page"},{"oldId":"b-63690cced36e0062c372afd8","targetId":"b-63690cced36e0062c372afd8","page":"reading-opening","operation":"move","beforeSHA256":"93743e4257248bf9219358b363aae08572036e7ccafb37b6917bd45eeacfbe96","afterSHA256":"93743e4257248bf9219358b363aae08572036e7ccafb37b6917bd45eeacfbe96","reason":"Same source body in authored semantic page"},{"oldId":"b-5dca42862f329c2bdd7d756e","targetId":"b-5dca42862f329c2bdd7d756e","page":"reading-opening","operation":"move","beforeSHA256":"ad5c18536ebaaed1683a9cbbc42e20e2f10a94f3ad582f81270f89de1719a98f","afterSHA256":"ad5c18536ebaaed1683a9cbbc42e20e2f10a94f3ad582f81270f89de1719a98f","reason":"Same source body in authored semantic page"},{"oldId":"b-569015fd6fb4f8e337fd406f","targetId":"b-569015fd6fb4f8e337fd406f","page":"reading-opening","operation":"move","beforeSHA256":"5409b0105bb5ec9dc3241c580e59c54772c6e8b9d8dbcfe5caf959343ea238a6","afterSHA256":"5409b0105bb5ec9dc3241c580e59c54772c6e8b9d8dbcfe5caf959343ea238a6","reason":"Same source body in authored semantic page"},{"oldId":"b-9d4c17635ddd9b4d360243bd","targetId":"b-9d4c17635ddd9b4d360243bd","page":"reading-opening","operation":"move","beforeSHA256":"77f6d53302f1be4d6d3ee57e02b21459f56e0aa208d6bbbd81accd9d8b377d53","afterSHA256":"77f6d53302f1be4d6d3ee57e02b21459f56e0aa208d6bbbd81accd9d8b377d53","reason":"Same source body in authored semantic page"},{"oldId":"b-31073228d6c1d8085620da2b","targetId":"b-31073228d6c1d8085620da2b","page":"reading-opening","operation":"move","beforeSHA256":"3e56a076a7b9a48f3f8c0e4a1ec3518b7589369f6099f713d3ab613354f864c0","afterSHA256":"3e56a076a7b9a48f3f8c0e4a1ec3518b7589369f6099f713d3ab613354f864c0","reason":"Same source body in authored semantic page"},{"oldId":"b-4f6c0d063152fd542a6f927f","targetId":"b-4f6c0d063152fd542a6f927f","page":"reading-opening","operation":"move","beforeSHA256":"ad790b051493beed5374493af79f9514dc8aa5e0b22af3b3dc4d840d31e23fe3","afterSHA256":"ad790b051493beed5374493af79f9514dc8aa5e0b22af3b3dc4d840d31e23fe3","reason":"Same source body in authored semantic page"},{"oldId":"b-53bdd6902874cc754662b94d","targetId":"b-53bdd6902874cc754662b94d","page":"reading-opening","operation":"move","beforeSHA256":"be0aaf3d04041ca7d223159fab6db6c9adaac7193b5a1788872d4d17ef5a188b","afterSHA256":"be0aaf3d04041ca7d223159fab6db6c9adaac7193b5a1788872d4d17ef5a188b","reason":"Same source body in authored semantic page"},{"oldId":"b-307a5d42ba91cb4cfa1d2e90","targetId":"b-307a5d42ba91cb4cfa1d2e90","page":"reading-opening","operation":"move","beforeSHA256":"a9149806afd622bccc16835a043e5f4608c2cf76dcabe0e7596a0cc5c2bf8c51","afterSHA256":"a9149806afd622bccc16835a043e5f4608c2cf76dcabe0e7596a0cc5c2bf8c51","reason":"Same source body in authored semantic page"},{"oldId":"b-fa200bbe470b7de23da9aad3","targetId":"b-fa200bbe470b7de23da9aad3","page":"reading-opening","operation":"move","beforeSHA256":"241bf275da03aa8d9391b79ee6cd44b57a829f6baa38512357bc65f453f808df","afterSHA256":"241bf275da03aa8d9391b79ee6cd44b57a829f6baa38512357bc65f453f808df","reason":"Same source body in authored semantic page"},{"oldId":"b-ee04472871fdfc878c288c60","targetId":"b-ee04472871fdfc878c288c60","page":"reading-opening","operation":"move","beforeSHA256":"cf7413ba356779e875fe91f4f34c7c315f7f7b7ce42e19b44ffd1dcb30d0d58b","afterSHA256":"cf7413ba356779e875fe91f4f34c7c315f7f7b7ce42e19b44ffd1dcb30d0d58b","reason":"Same source body in authored semantic page"},{"oldId":"b-2721e5c699bbff48b1e9a5d9","targetId":"b-2721e5c699bbff48b1e9a5d9","page":"reading-opening","operation":"move","beforeSHA256":"041babd888393a13f260ef325bcc31c4c9cd1e4f7df2491e324ddafb2b56cd23","afterSHA256":"041babd888393a13f260ef325bcc31c4c9cd1e4f7df2491e324ddafb2b56cd23","reason":"Same source body in authored semantic page"},{"oldId":"b-973a60e5afd310d4446eaeab","targetId":"b-973a60e5afd310d4446eaeab","page":"reading-opening","operation":"move","beforeSHA256":"c315ee35160727db2c7bb7bef724e45266800f78ea5d2880c7d3101479ad3b43","afterSHA256":"c315ee35160727db2c7bb7bef724e45266800f78ea5d2880c7d3101479ad3b43","reason":"Same source body in authored semantic page"},{"oldId":"b-f41b9064bce33bc54a864235","targetId":"b-f41b9064bce33bc54a864235","page":"reading-opening","operation":"move","beforeSHA256":"9d6d23d8e3c027ed4f0115e54b8a2a363b6e2f8a619951c7433c03f79bcd0992","afterSHA256":"9d6d23d8e3c027ed4f0115e54b8a2a363b6e2f8a619951c7433c03f79bcd0992","reason":"Same source body in authored semantic page"},{"oldId":"b-8e5e23939608a12d622a4b63","targetId":"b-8e5e23939608a12d622a4b63","page":"reading-opening","operation":"move","beforeSHA256":"ae393bf00717f2b8ab32d920a04af01319f2e21853fbca2d696f4bb48f770610","afterSHA256":"ae393bf00717f2b8ab32d920a04af01319f2e21853fbca2d696f4bb48f770610","reason":"Same source body in authored semantic page"},{"oldId":"b-277ff9ac1fa96bbde6a03248","targetId":"b-277ff9ac1fa96bbde6a03248","page":"reading-opening","operation":"move","beforeSHA256":"35d7aa5a6598e52321e9309a0916b06dc03cd25c5ad4df6383272b464361299a","afterSHA256":"35d7aa5a6598e52321e9309a0916b06dc03cd25c5ad4df6383272b464361299a","reason":"Same source body in authored semantic page"},{"oldId":"b-8e0df0244f6f4651f3ff0566","targetId":"b-8e0df0244f6f4651f3ff0566","page":"reading-opening","operation":"move","beforeSHA256":"cdf5e1ee80a89ea700fef3561ecdfa2d3713ae806370d2924e5897e2a5078529","afterSHA256":"cdf5e1ee80a89ea700fef3561ecdfa2d3713ae806370d2924e5897e2a5078529","reason":"Same source body in authored semantic page"},{"oldId":"b-5c9e07f83740e2ba8d6648bc","targetId":"b-5c9e07f83740e2ba8d6648bc","page":"reading-opening","operation":"move","beforeSHA256":"d8c17ff28b5b0d5b1671238ddcef7a1c477eeded2b48b525dadaddb86200d640","afterSHA256":"d8c17ff28b5b0d5b1671238ddcef7a1c477eeded2b48b525dadaddb86200d640","reason":"Same source body in authored semantic page"},{"oldId":"b-9d63845e5b2c1fe018bef75b","targetId":"b-9d63845e5b2c1fe018bef75b","page":"reading-opening","operation":"move","beforeSHA256":"13df4d569b221f8bc60c5de960a783d1972c800851f3e5ebd251ca5ca31bb27c","afterSHA256":"13df4d569b221f8bc60c5de960a783d1972c800851f3e5ebd251ca5ca31bb27c","reason":"Same source body in authored semantic page"},{"oldId":"b-4235d02d8f7094174f01c76f","targetId":"b-4235d02d8f7094174f01c76f","page":"reading-opening","operation":"rewrite","beforeSHA256":"5530957dfc389313134c1ae2e8fbba7cb2ec21f5033903c548698fd1182a0ee5","afterSHA256":"d12b72281bce9d27865a10e904405cca1c2c915b9910c63934ad6f7d4439e49a","reason":"Same source identity with figure semantics or reading context enriched"},{"oldId":"b-9f087043252dc6d1d7e05e56","targetId":"b-9f087043252dc6d1d7e05e56","page":"reading-opening","operation":"move","beforeSHA256":"4c2978607bb7b437fdad1528eb065e99e5968fdc50592f9201ed43dcf3ded1a3","afterSHA256":"4c2978607bb7b437fdad1528eb065e99e5968fdc50592f9201ed43dcf3ded1a3","reason":"Same source body in authored semantic page"},{"oldId":"b-5f0511ef256ae30e0fb49904","targetId":"b-5f0511ef256ae30e0fb49904","page":"reading-opening","operation":"move","beforeSHA256":"939045e192e80393bc29ee5638c03a5341285cb4e18440f27403c32464cc6f54","afterSHA256":"939045e192e80393bc29ee5638c03a5341285cb4e18440f27403c32464cc6f54","reason":"Same source body in authored semantic page"},{"oldId":"b-35677dacae84318488d3a879","targetId":"b-35677dacae84318488d3a879","page":"reading-opening","operation":"move","beforeSHA256":"715899b38a78b3bf4f5526299aecfd4f7c5178ff28d2193adb58e1166be1ad9d","afterSHA256":"715899b38a78b3bf4f5526299aecfd4f7c5178ff28d2193adb58e1166be1ad9d","reason":"Same source body in authored semantic page"},{"oldId":"b-6556b03947d3dbfb53faeabe","targetId":"b-6556b03947d3dbfb53faeabe","page":"reading-opening","operation":"rewrite","beforeSHA256":"1bdd695c84d8c10d3518a5f111bf260a1bf6eb69d2b1d291cf172813e2d66865","afterSHA256":"506f6e140199a9480864e32072bf7dc411386e1f8072f90aaf04d4a7616bf87e","reason":"Same source identity with figure semantics or reading context enriched"},{"oldId":"b-7975ac05ef7e78d12b9a0d80","targetId":"b-7975ac05ef7e78d12b9a0d80","page":"reading-opening","operation":"move","beforeSHA256":"596f845a43d7a997861347a6bbdfe900ef177b21daf6e5f5bc7a191ee6b9e7f8","afterSHA256":"596f845a43d7a997861347a6bbdfe900ef177b21daf6e5f5bc7a191ee6b9e7f8","reason":"Same source body in authored semantic page"},{"oldId":"b-99cee38ffedf0720c9c26c3a","targetId":"b-99cee38ffedf0720c9c26c3a","page":"reading-opening","operation":"move","beforeSHA256":"3692ed029bf31213ee6ddce712d05db3ed57c54d18122d873965128ee09e9c12","afterSHA256":"f56f575033bcab034622fd5bcf423ca70f6f99a0671bfa0c2557bc4ade38d310","reason":"Same source body in authored semantic page"},{"oldId":"b-0e425b3f42d468b14853afc8","targetId":"b-0e425b3f42d468b14853afc8","page":"reading-opening","operation":"move","beforeSHA256":"0bc30a0c85f509f96425ef6d503b1f4dda54adc0ee2017a6d11f91fd6398d7a1","afterSHA256":"6e59fe129adbd7f631cd6517b8715f8e88361fc81316790d51a971ff6a7ac9d1","reason":"Same source body in authored semantic page"},{"oldId":"b-767d6f3ae456a8a61c723243","targetId":"b-767d6f3ae456a8a61c723243","page":"reading-opening","operation":"move","beforeSHA256":"239d41a1854eaf46c28660b8f1fafbd9e2dd5243ba3e50949f8b9548fb79b67f","afterSHA256":"239d41a1854eaf46c28660b8f1fafbd9e2dd5243ba3e50949f8b9548fb79b67f","reason":"Same source body in authored semantic page"},{"oldId":"block-6ca34c15abaf4aeb9900af72759ebc28--一张类别图怎样成为可训练的输出","targetId":"b-9d4c17635ddd9b4d360243bd","operation":"retain-source-identity"},{"oldId":"block-6ca34c15abaf4aeb9900af72759ebc28--上采样能增加格子不能单凭格子恢复被删掉的区别","targetId":"b-fa200bbe470b7de23da9aad3","operation":"retain-source-identity"},{"oldId":"block-6ca34c15abaf4aeb9900af72759ebc28--一个-88-反例怎样澄清四种指标","targetId":"b-277ff9ac1fa96bbde6a03248","operation":"retain-source-identity"}],"blocks":[{"id":"b-f3dc1ec29e18aec577e14dda","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-f3dc1ec29e18aec577e14dda","content":"[CLS 08 · 密集预测、实例与集合输出](/course/notebooks/2d81e2a1936043349eacb71c53c0defa.html)","sha256":"83aaf240c580bb5e1a3c458920748ed89991c349313e733e2c8bb2d5ae6ac25d","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-8a65403a37d4bbd2d691096b","prerequisites":[],"relatedBlocks":[]},{"id":"b-8a65403a37d4bbd2d691096b","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-8a65403a37d4bbd2d691096b","content":"# 一张类别图怎样成为可训练的输出","sha256":"6afaf53c5c378b64998573e505c11a4a5921e1c2b4ec5292171acc7ce6507a3c","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-f3dc1ec29e18aec577e14dda","nextBlock":"b-0863ea013ba4265d69619fab","prerequisites":[{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-0863ea013ba4265d69619fab","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-0863ea013ba4265d69619fab","content":"按问题继续","sha256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-8a65403a37d4bbd2d691096b","nextBlock":"b-c561639ba6edfdbb0eb7aee8","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-c561639ba6edfdbb0eb7aee8","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-c561639ba6edfdbb0eb7aee8","content":"像素损失、IoU口径、粗细表示、实例身份与集合输出","sha256":"70eec56f79b131d4695a1c420bc695a02ef3863727481994de8f7ed4bd673dcc","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-0863ea013ba4265d69619fab","nextBlock":"b-802b09f0ef7cd549038d90a7","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-802b09f0ef7cd549038d90a7","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-802b09f0ef7cd549038d90a7","content":"**学习路径：** [像素、边界与对象身份](/course/study/pixels-instances.html)","sha256":"30ab9f304366670818536cf73c98c66aa814e8e31ecdbffcf19a4d6e7e7f22de","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-c561639ba6edfdbb0eb7aee8","nextBlock":"b-29e04a304ea18f0cb8f0f6bd","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-29e04a304ea18f0cb8f0f6bd","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-29e04a304ea18f0cb8f0f6bd","content":"本册承接 [P36 的逐像素任务](/course/pages/l3-r2-p36.html)，解释 [LAB 11](https://deepnote.com/project/38755e4e-23f6-4f5f-8c68-8a2e4114b1c1/notebook/9e268706471944e3a59e6f568f9ce9a3) 为什么需要新的目标、损失和评价。先读 logits/目标/ignore，再用 8×8 算例区分前景 IoU、两类 mIoU 和像素准确率；最后看语义类别为何不能唯一给出实例身份。","sha256":"72798cbfc37e71e8b9dbbcfe3cfb8a7b0fc2d74531b251cdcb3e0c9989b9c37d","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-802b09f0ef7cd549038d90a7","nextBlock":"b-63690cced36e0062c372afd8","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-63690cced36e0062c372afd8","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-63690cced36e0062c372afd8","content":"希望先看结果，可读 [RGB 读出与 FCN 的对照](/course/blocks/pixel-model-positive-result.html)；希望核一个指标，可引用 [0.667 与 0.75 的完整算例](/course/blocks/mask-iou-worked-example.html)。读完应能解释“框对了但边界仍错”和“所有杯子像素对了但实例仍未分开”这两个缺口。","sha256":"93743e4257248bf9219358b363aae08572036e7ccafb37b6917bd45eeacfbe96","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-29e04a304ea18f0cb8f0f6bd","nextBlock":"b-5dca42862f329c2bdd7d756e","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5dca42862f329c2bdd7d756e","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-5dca42862f329c2bdd7d756e","content":"坐标采样的动手入口是 [LAB 03](https://deepnote.com/project/38755e4e-23f6-4f5f-8c68-8a2e4114b1c1/notebook/49a2c030eafa4e1db6fc4f1a08c3810a)；多尺度和集合匹配留在本册后部选读。它们不要求在理解像素预测前先完成。","sha256":"ad5c18536ebaaed1683a9cbbc42e20e2f10a94f3ad582f81270f89de1719a98f","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-63690cced36e0062c372afd8","nextBlock":"b-569015fd6fb4f8e337fd406f","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-569015fd6fb4f8e337fd406f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-569015fd6fb4f8e337fd406f","content":"一个框可以同时包住杯身、杯把和杯把中的背景。框坐标全对，仍不代表我们知道每个像素属于谁。若两只杯子接触，语义分割可以把它们都标成“杯子”，但要分别计数和操作，就还需区分实例身份。这里不是同一个答案换显示颜色，而是输出空间和监督单位发生了变化。","sha256":"5409b0105bb5ec9dc3241c580e59c54772c6e8b9d8dbcfe5caf959343ea238a6","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-5dca42862f329c2bdd7d756e","nextBlock":"b-9d4c17635ddd9b4d360243bd","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-9d4c17635ddd9b4d360243bd","type":"h2","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-9d4c17635ddd9b4d360243bd","content":"## 一张类别图怎样成为可训练的输出","sha256":"77f6d53302f1be4d6d3ee57e02b21459f56e0aa208d6bbbd81accd9d8b377d53","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-569015fd6fb4f8e337fd406f","nextBlock":"b-31073228d6c1d8085620da2b","prerequisites":[{"id":"b-8a65403a37d4bbd2d691096b","url":"#b-8a65403a37d4bbd2d691096b","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-31073228d6c1d8085620da2b","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-31073228d6c1d8085620da2b","content":"语义分割的输出通常为 $Z\\in\\mathbb R^{N\\times C\\times H\\times W}$，其中 $N$ 是 batch，$C$ 是类别数，$H,W$ 是空间大小。目标 $Y\\in\\{0,\\ldots,C-1\\}^{N\\times H\\times W}$ 是整数类别图。对每个有效像素做 softmax 交叉熵，再按有效像素数平均：","sha256":"3e56a076a7b9a48f3f8c0e4a1ec3518b7589369f6099f713d3ab613354f864c0","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"Z\\in\\mathbb R^{N\\times C\\times H\\times W}","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"N","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"C","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"H,W","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"Y\\in\\{0,\\ldots,C-1\\}^{N\\times H\\times W}","definitionContext":"一张类别图怎样成为可训练的输出"}],"previousBlock":"b-9d4c17635ddd9b4d360243bd","nextBlock":"b-4f6c0d063152fd542a6f927f","prerequisites":[{"id":"b-9d4c17635ddd9b4d360243bd","url":"#b-9d4c17635ddd9b4d360243bd","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-4f6c0d063152fd542a6f927f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-4f6c0d063152fd542a6f927f","content":"$L=-\\frac{1}{|\\Omega|}\\sum_{(n,u,v)\\in\\Omega}\\log p_{n,Y_{nuv},u,v}.$","sha256":"ad790b051493beed5374493af79f9514dc8aa5e0b22af3b3dc4d840d31e23fe3","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"L=-\\frac{1}{|\\Omega|}\\sum_{(n,u,v)\\in\\Omega}\\log p_{n,Y_{nuv},u,v}.","definitionContext":"一张类别图怎样成为可训练的输出"}],"previousBlock":"b-31073228d6c1d8085620da2b","nextBlock":"b-53bdd6902874cc754662b94d","prerequisites":[{"id":"b-9d4c17635ddd9b4d360243bd","url":"#b-9d4c17635ddd9b4d360243bd","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-53bdd6902874cc754662b94d","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-53bdd6902874cc754662b94d","content":"$\\Omega$ 排除 ignore label，例如未标注区域。ignore 不应变成一个额外背景类，也不进入评价混淆矩阵。二类任务可以用一个 foreground logit 加 sigmoid/BCE，也可以用两个 logits 加 softmax；张量形状和损失接口必须与这种选择一致。","sha256":"be0aaf3d04041ca7d223159fab6db6c9adaac7193b5a1788872d4d17ef5a188b","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"\\Omega","definitionContext":"一张类别图怎样成为可训练的输出"}],"previousBlock":"b-4f6c0d063152fd542a6f927f","nextBlock":"b-307a5d42ba91cb4cfa1d2e90","prerequisites":[{"id":"b-9d4c17635ddd9b4d360243bd","url":"#b-9d4c17635ddd9b4d360243bd","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-307a5d42ba91cb4cfa1d2e90","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-307a5d42ba91cb4cfa1d2e90","content":"FCN 用卷积保持空间网格并输出位置相关分数，再把较粗分辨率恢复到目标大小。U-Net 的编码—解码结构把深层上下文与浅层高分辨率特征通过跳连结合；这种连接提供额外可用信息，而不仅是把同一张粗图放大。[D2L §14.11 FCN](https://d2l.ai/chapter_computer-vision/fcn.html) · [U-Net 原论文](https://arxiv.org/abs/1505.04597)。图像和 mask 的裁剪、翻转、缩放必须使用同一几何变换，否则网络被要求把一个位置的像素对应到另一个位置的标签。","sha256":"a9149806afd622bccc16835a043e5f4608c2cf76dcabe0e7596a0cc5c2bf8c51","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-53bdd6902874cc754662b94d","nextBlock":"b-fa200bbe470b7de23da9aad3","prerequisites":[{"id":"b-9d4c17635ddd9b4d360243bd","url":"#b-9d4c17635ddd9b4d360243bd","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-fa200bbe470b7de23da9aad3","type":"h2","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-fa200bbe470b7de23da9aad3","content":"## 上采样能增加格子，不能单凭格子恢复被删掉的区别","sha256":"241bf275da03aa8d9391b79ee6cd44b57a829f6baa38512357bc65f453f808df","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-307a5d42ba91cb4cfa1d2e90","nextBlock":"b-ee04472871fdfc878c288c60","prerequisites":[{"id":"b-9d4c17635ddd9b4d360243bd","url":"#b-9d4c17635ddd9b4d360243bd","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-ee04472871fdfc878c288c60","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-ee04472871fdfc878c288c60","content":"最近邻复制相邻值；双线性插值对邻域数值加权。给二维网格","sha256":"cf7413ba356779e875fe91f4f34c7c315f7f7b7ce42e19b44ffd1dcb30d0d58b","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-fa200bbe470b7de23da9aad3","nextBlock":"b-2721e5c699bbff48b1e9a5d9","prerequisites":[{"id":"b-fa200bbe470b7de23da9aad3","url":"#b-fa200bbe470b7de23da9aad3","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-2721e5c699bbff48b1e9a5d9","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-2721e5c699bbff48b1e9a5d9","content":"$F=\\begin{pmatrix}0&10\\\\20&30\\end{pmatrix},$","sha256":"041babd888393a13f260ef325bcc31c4c9cd1e4f7df2491e324ddafb2b56cd23","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"F=\\begin{pmatrix}0&10\\\\20&30\\end{pmatrix},","definitionContext":"一张类别图怎样成为可训练的输出"}],"previousBlock":"b-ee04472871fdfc878c288c60","nextBlock":"b-973a60e5afd310d4446eaeab","prerequisites":[{"id":"b-fa200bbe470b7de23da9aad3","url":"#b-fa200bbe470b7de23da9aad3","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-973a60e5afd310d4446eaeab","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-973a60e5afd310d4446eaeab","content":"在横坐标 $x=0.25$、纵坐标 $y=0.5$ 处，四点权重为 $0.375,0.125,0.375,0.125$，插值得到 12.5。它精确说明怎样从已有数值取连续位置，不能保证 12.5 就是原始图像那一点的真实内容。","sha256":"c315ee35160727db2c7bb7bef724e45266800f78ea5d2880c7d3101479ad3b43","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"x=0.25","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"y=0.5","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"0.375,0.125,0.375,0.125","definitionContext":"一张类别图怎样成为可训练的输出"}],"previousBlock":"b-2721e5c699bbff48b1e9a5d9","nextBlock":"b-f41b9064bce33bc54a864235","prerequisites":[{"id":"b-fa200bbe470b7de23da9aad3","url":"#b-fa200bbe470b7de23da9aad3","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-f41b9064bce33bc54a864235","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-f41b9064bce33bc54a864235","content":"如果两个 $2\\times2$ 图案“左半亮”和“右半亮”都被平均成 0.5，任何只拿这个平均值的确定性解码器都必须给二者相同输出。增加一个“左减右”通道后才可以区分。这是 [LAB 03](https://deepnote.com/workspace/ise3309jingao-5ba66b0a-4283-4f5a-ae06-3f68d3201e99/project/38755e4e-23f6-4f5f-8c68-8a2e4114b1c1/notebook/49a2c030eafa4e1db6fc4f1a08c3810a?utm_source=openai&utm_medium=mcp&utm_campaign=openaimcp&utm_content=49a2c030eafa4e1db6fc4f1a08c3810a&utm_term=get_notebook) 的表示碰撞例；它解释为什么需要保留空间信息，而非只提高显示分辨率。","sha256":"9d6d23d8e3c027ed4f0115e54b8a2a363b6e2f8a619951c7433c03f79bcd0992","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"2\\times2","definitionContext":"一张类别图怎样成为可训练的输出"}],"previousBlock":"b-973a60e5afd310d4446eaeab","nextBlock":"b-8e5e23939608a12d622a4b63","prerequisites":[{"id":"b-fa200bbe470b7de23da9aad3","url":"#b-fa200bbe470b7de23da9aad3","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-8e5e23939608a12d622a4b63","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-8e5e23939608a12d622a4b63","content":"类别 mask 本身是离散 ID，通常用最近邻几何采样。对类别编号 1 和 3 做双线性插值，得到 2 并不意味着中间真的存在第 2 类。模型的连续 logits 或概率图则可以按适当约定插值，再 argmax/阈值。插值坐标约定也要固定，例如是否对齐端点；同样叫 bilinear 的两个设置可能在边界位置不同。","sha256":"ae393bf00717f2b8ab32d920a04af01319f2e21853fbca2d696f4bb48f770610","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-f41b9064bce33bc54a864235","nextBlock":"b-277ff9ac1fa96bbde6a03248","prerequisites":[{"id":"b-fa200bbe470b7de23da9aad3","url":"#b-fa200bbe470b7de23da9aad3","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-277ff9ac1fa96bbde6a03248","type":"h2","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-277ff9ac1fa96bbde6a03248","content":"## 一个 8×8 反例怎样澄清四种指标","sha256":"35d7aa5a6598e52321e9309a0916b06dc03cd25c5ad4df6383272b464361299a","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-8e5e23939608a12d622a4b63","nextBlock":"b-8e0df0244f6f4651f3ff0566","prerequisites":[{"id":"b-fa200bbe470b7de23da9aad3","url":"#b-fa200bbe470b7de23da9aad3","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-8e0df0244f6f4651f3ff0566","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-8e0df0244f6f4651f3ff0566","content":"设真值前景有 16 像素，从真值 mask 平均降采样再最近邻恢复并阈值化后，预测前景有 24 像素，其中真阳性 16、假阳性 8、假阴性 0、真阴性 40。这是采样机制例，预测过程直接使用了真值 mask，不能当作看图模型的能力证据。","sha256":"cdf5e1ee80a89ea700fef3561ecdfa2d3713ae806370d2924e5897e2a5078529","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-277ff9ac1fa96bbde6a03248","nextBlock":"b-5c9e07f83740e2ba8d6648bc","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5c9e07f83740e2ba8d6648bc","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-5c9e07f83740e2ba8d6648bc","content":"![真值掩码、降采样恢复和错误区域](/course/repair/assets/cls09-mask-metrics.svg)","sha256":"d8c17ff28b5b0d5b1671238ddcef7a1c477eeded2b48b525dadaddb86200d640","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-8e0df0244f6f4651f3ff0566","nextBlock":"b-9d63845e5b2c1fe018bef75b","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-9d63845e5b2c1fe018bef75b","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-9d63845e5b2c1fe018bef75b","content":"三张网格由同一组 8×8 数组绘制。目标矩形为行 2–5、列 3–6（从 0 计数）；阈值化最近邻恢复多出了 8 个前景像素。","sha256":"13df4d569b221f8bc60c5de960a783d1972c800851f3e5ebd251ca5ca31bb27c","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-5c9e07f83740e2ba8d6648bc","nextBlock":"b-4235d02d8f7094174f01c76f","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-4235d02d8f7094174f01c76f","type":"table","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-4235d02d8f7094174f01c76f","content":"| 指标 | 计算 | 本例结果 |\n| --- | --- | --- |\n| 前景 IoU | $16/(16+8+0)$ | $2/3$ |\n| 背景 IoU | $40/(40+0+8)$ | $5/6$ |\n| 两类 mIoU，含背景 | $(2/3+5/6)/2$ | 0.75 |\n| 像素准确率 | $(16+40)/64$ | 0.875 |","sha256":"d12b72281bce9d27865a10e904405cca1c2c915b9910c63934ad6f7d4439e49a","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"16/(16+8+0)","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"2/3","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"40/(40+0+8)","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"5/6","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"(2/3+5/6)/2","definitionContext":"一张类别图怎样成为可训练的输出"},{"latex":"(16+40)/64","definitionContext":"一张类别图怎样成为可训练的输出"}],"table":{"headers":["指标","计算","本例结果"],"rows":[["指标","计算","本例结果"],["前景 IoU","16/(16+8+0)16/(16+8+0)16/(16+8+0)","2/32/32/3"],["背景 IoU","40/(40+0+8)40/(40+0+8)40/(40+0+8)","5/65/65/6"],["两类 mIoU，含背景","(2/3+5/6)/2(2/3+5/6)/2(2/3+5/6)/2","0.75"],["像素准确率","(16+40)/64(16+40)/64(16+40)/64","0.875"]]},"previousBlock":"b-9d63845e5b2c1fe018bef75b","nextBlock":"b-9f087043252dc6d1d7e05e56","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-9f087043252dc6d1d7e05e56","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-9f087043252dc6d1d7e05e56","content":"“mIoU=0.667”在这个例子里错在名称：0.667 是单类前景 IoU。若报告不含背景的类均值，应写清类别集合；若某类真值和预测都为空，要说明是忽略该类还是采用约定值。背景占比很高时，全预测背景也可能有高像素准确率，所以要看逐类混淆和边界错误，而不只看一个总数。","sha256":"4c2978607bb7b437fdad1528eb065e99e5968fdc50592f9201ed43dcf3ded1a3","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-4235d02d8f7094174f01c76f","nextBlock":"b-5f0511ef256ae30e0fb49904","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5f0511ef256ae30e0fb49904","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-5f0511ef256ae30e0fb49904","content":"真正视觉模型必须从图像推断 mask。[LAB 11](https://deepnote.com/workspace/ise3309jingao-5ba66b0a-4283-4f5a-ae06-3f68d3201e99/project/38755e4e-23f6-4f5f-8c68-8a2e4114b1c1/notebook/9e268706471944e3a59e6f568f9ce9a3?utm_source=openai&utm_medium=mcp&utm_campaign=openaimcp&utm_content=9e268706471944e3a59e6f568f9ce9a3&utm_term=get_notebook) 先保留上述采样例及最近邻/双线性比较，再进入小型像素模型。两段实验分别检验采样损失和学习能力，结论不互相替代。","sha256":"939045e192e80393bc29ee5638c03a5341285cb4e18440f27403c32464cc6f54","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-9f087043252dc6d1d7e05e56","nextBlock":"b-35677dacae84318488d3a879","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-35677dacae84318488d3a879","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-35677dacae84318488d3a879","content":"主实验在 2026-09-22 本地 CPU 完成训练，384 张训练图、128 张保留图，输入 48×48 RGB。三类为背景、红矩形、蓝椭圆，目标形状为 `[B,48,48]`，logits 为 `[B,3,48,48]`，ignore label 255 在损失与指标中均被排除。20,590 参数的小型 FCN 将 12×12 粗 logits 双线性上采样并融合精细跳连，训练 20 epoch；另训练只有 12 个参数的逐像素 RGB 线性读出作基线。","sha256":"715899b38a78b3bf4f5526299aecfd4f7c5178ff28d2193adb58e1166be1ad9d","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-5f0511ef256ae30e0fb49904","nextBlock":"b-6556b03947d3dbfb53faeabe","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-6556b03947d3dbfb53faeabe","type":"table","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-6556b03947d3dbfb53faeabe","content":"| 保留图像条件 | 小型 FCN 三类 mIoU | RGB 逐像素线性读出 mIoU |\n| --- | --- | --- |\n| 原始合成图 | 0.99041 | 1.00000 |\n| 对象尺度变为 0.45 倍 | 0.94396 | 1.00000 |\n| RGB 图作 5×5 模糊 | 0.92397 | 0.84996 |","sha256":"506f6e140199a9480864e32072bf7dc411386e1f8072f90aaf04d4a7616bf87e","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"table":{"headers":["保留图像条件","小型 FCN 三类 mIoU","RGB 逐像素线性读出 mIoU"],"rows":[["保留图像条件","小型 FCN 三类 mIoU","RGB 逐像素线性读出 mIoU"],["原始合成图","0.99041","1.00000"],["对象尺度变为 0.45 倍","0.94396","1.00000"],["RGB 图作 5×5 模糊","0.92397","0.84996"]]},"previousBlock":"b-35677dacae84318488d3a879","nextBlock":"b-7975ac05ef7e78d12b9a0d80","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-7975ac05ef7e78d12b9a0d80","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-7975ac05ef7e78d12b9a0d80","content":"RGB 基线满分说明原数据颜色已直接泄露类别，不能据 FCN 高分宣称它学会了复杂形状或需要深层上下文。模糊后边缘颜色混合，FCN 在这个干预下更稳健；这才是当前对照支持的具体比较。原始条件下 FCN 的逐类 IoU 为背景 0.99861、矩形 0.99833、椭圆 0.97429，较细曲线边界更难。","sha256":"596f845a43d7a997861347a6bbdfe900ef177b21daf6e5f5bc7a191ee6b9e7f8","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6556b03947d3dbfb53faeabe","nextBlock":"b-99cee38ffedf0720c9c26c3a","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-99cee38ffedf0720c9c26c3a","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-99cee38ffedf0720c9c26c3a","content":"![真实图像、mask、FCN输出与错误](/assets/display/f2028b85a049b53efbcbbb7a.webp)","sha256":"f56f575033bcab034622fd5bcf423ca70f6f99a0671bfa0c2557bc4ade38d310","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-7975ac05ef7e78d12b9a0d80","nextBlock":"b-0e425b3f42d468b14853afc8","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-0e425b3f42d468b14853afc8","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-0e425b3f42d468b14853afc8","content":"![小对象条件下的错误分布](/assets/display/a3b94d7df0f7bd4472d4f87d.webp)","sha256":"6e59fe129adbd7f631cd6517b8715f8e88361fc81316790d51a971ff6a7ac9d1","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-99cee38ffedf0720c9c26c3a","nextBlock":"b-767d6f3ae456a8a61c723243","prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-767d6f3ae456a8a61c723243","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28.html#b-767d6f3ae456a8a61c723243","content":"保持已训练参数，测试时仅把粗 logits 的双线性插值改为最近邻，mIoU 降到 0.91180。这说明当前网络对训练时的采样约定敏感，不能把它读作“最近邻永远更差”，因为网络没有在两种上采样设置下分别重训。前面的已知 mask 热身中，双线性分支前景 IoU 仍为 2/3，背景 IoU 为 0.88462、两类 mIoU 为 0.77564；指标差异与模型泛化仍须分开。","sha256":"239d41a1854eaf46c28660b8f1fafbd9e2dd5243ba3e50949f8b9548fb79b67f","chapter":{"id":"course","title":"一张类别图怎样成为可训练的输出"},"unit":{"id":"notebook-blocks/6ca34c15abaf4aeb9900af72759ebc28","title":"一张类别图怎样成为可训练的输出"},"page":{"id":"reading-opening","title":"一张类别图怎样成为可训练的输出","number":1},"sectionTitle":"一张类别图怎样成为可训练的输出","currentQuestion":"一张类别图怎样成为可训练的输出","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-0e425b3f42d468b14853afc8","nextBlock":null,"prerequisites":[{"id":"b-277ff9ac1fa96bbde6a03248","url":"#b-277ff9ac1fa96bbde6a03248","role":"section-title"},{"id":"b-f3dc1ec29e18aec577e14dda","url":"#b-f3dc1ec29e18aec577e14dda","role":"problem-context"}],"relatedBlocks":[]}]}