{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html","markdown":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.md","pages":[{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","unit":"一句话怎样落实为正确对象的区域？","unitId":"grounding-object-contract","chapter":"III","number":1}],"aliases":[{"oldId":"b-a06529c46ba769c1bc0f0e50","targetId":"b-a06529c46ba769c1bc0f0e50","page":"reading-opening","operation":"move","beforeSHA256":"a33f5de7b37ed260eed9f501944f3676e0a2020d6d0f6a871e4b2c81fa56ff43","afterSHA256":"a33f5de7b37ed260eed9f501944f3676e0a2020d6d0f6a871e4b2c81fa56ff43","reason":"Same source body in authored semantic page"},{"oldId":"b-4ce3918203caf3eefd2abef9","targetId":"b-4ce3918203caf3eefd2abef9","page":"reading-opening","operation":"move","beforeSHA256":"7d8bc1408f3f30ddf3278f32dbb54c475cf47f3b9b5ab00d4d7f1ef8f718627d","afterSHA256":"7d8bc1408f3f30ddf3278f32dbb54c475cf47f3b9b5ab00d4d7f1ef8f718627d","reason":"Same source body in authored semantic page"},{"oldId":"b-e7dd1dd7768c13f1ecb98165","targetId":"b-e7dd1dd7768c13f1ecb98165","page":"reading-opening","operation":"move","beforeSHA256":"ecc2d4cc35754ab1239dd38d2bcab0d3de0a2b3e1a2f7a7a6055bad2349326fd","afterSHA256":"ecc2d4cc35754ab1239dd38d2bcab0d3de0a2b3e1a2f7a7a6055bad2349326fd","reason":"Same source body in authored semantic page"},{"oldId":"b-60556010adce6f629d92884d","targetId":"b-60556010adce6f629d92884d","page":"reading-opening","operation":"move","beforeSHA256":"fa64c7e04210b5383ba8acd1b35e411888d7e97d955ce692b290f61d15fc651f","afterSHA256":"fa64c7e04210b5383ba8acd1b35e411888d7e97d955ce692b290f61d15fc651f","reason":"Same source body in authored semantic page"},{"oldId":"b-baccd3c4d98ebdcf709fff0c","targetId":"b-baccd3c4d98ebdcf709fff0c","page":"reading-opening","operation":"move","beforeSHA256":"a7e6186a3879cf00b7b4e6691c1a5fe72292766a0c2120f0667b376036a914b6","afterSHA256":"a7e6186a3879cf00b7b4e6691c1a5fe72292766a0c2120f0667b376036a914b6","reason":"Same source body in authored semantic page"},{"oldId":"b-7437a583e1091e603e0afafd","targetId":"b-7437a583e1091e603e0afafd","page":"reading-opening","operation":"move","beforeSHA256":"374d2bfb115c9901efd89d4182bd0a05ca1076977308b072a18ea0d33bcb6d4a","afterSHA256":"374d2bfb115c9901efd89d4182bd0a05ca1076977308b072a18ea0d33bcb6d4a","reason":"Same source body in authored semantic page"},{"oldId":"b-2f494a40f81c0af5249e317e","targetId":"b-2f494a40f81c0af5249e317e","page":"reading-opening","operation":"move","beforeSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","afterSHA256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-a06529c46ba769c1bc0f0e50","type":"h1","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-a06529c46ba769c1bc0f0e50","content":"# “蓝杯左边的杯子”：目标、参照物、关系三者分开","sha256":"a33f5de7b37ed260eed9f501944f3676e0a2020d6d0f6a871e4b2c81fa56ff43","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-4ce3918203caf3eefd2abef9","prerequisites":[{"id":"b-4ce3918203caf3eefd2abef9","url":"#b-4ce3918203caf3eefd2abef9","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-4ce3918203caf3eefd2abef9","type":"p","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-4ce3918203caf3eefd2abef9","content":"语言条件落实为空间输出时，需要明确目标、参照物和关系。GLIP 提供对象级学习和迁移的研究路线；LAB15 提供可逐像素追踪的规则实验。两者的层级不同，但都帮助判断系统到底把框放在谁身上，再与后续掩码处理连接。","sha256":"7d8bc1408f3f30ddf3278f32dbb54c475cf47f3b9b5ab00d4d7f1ef8f718627d","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-a06529c46ba769c1bc0f0e50","nextBlock":"b-e7dd1dd7768c13f1ecb98165","prerequisites":[{"id":"b-a06529c46ba769c1bc0f0e50","url":"#b-a06529c46ba769c1bc0f0e50","role":"section-title"}],"relatedBlocks":[]},{"id":"b-e7dd1dd7768c13f1ecb98165","type":"p","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-e7dd1dd7768c13f1ecb98165","content":"短语 `cup left of blue cup` 中，目标类别是 cup，参照物是 blue cup，关系是 left of。看到 blue 就返回蓝杯，找到了参照物，却没有完成目标选择。","sha256":"ecc2d4cc35754ab1239dd38d2bcab0d3de0a2b3e1a2f7a7a6055bad2349326fd","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-4ce3918203caf3eefd2abef9","nextBlock":"b-60556010adce6f629d92884d","prerequisites":[{"id":"b-a06529c46ba769c1bc0f0e50","url":"#b-a06529c46ba769c1bc0f0e50","role":"section-title"},{"id":"b-4ce3918203caf3eefd2abef9","url":"#b-4ce3918203caf3eefd2abef9","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-60556010adce6f629d92884d","type":"p","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-60556010adce6f629d92884d","content":"LAB15 的受控画布中，左红杯框为 $[15,19,50,58]$，中心横坐标 32.5；右蓝杯框为 $[109,19,144,58]$，中心横坐标 126.5；书的框为 $[60,65,90,80]$。本例定义 left-of 为目标中心横坐标小于参照中心，且目标不是参照物本身。","sha256":"fa64c7e04210b5383ba8acd1b35e411888d7e97d955ce692b290f61d15fc651f","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"[15,19,50,58]","definitionContext":"“蓝杯左边的杯子”：目标、参照物、关系三者分开"},{"latex":"[109,19,144,58]","definitionContext":"“蓝杯左边的杯子”：目标、参照物、关系三者分开"},{"latex":"[60,65,90,80]","definitionContext":"“蓝杯左边的杯子”：目标、参照物、关系三者分开"}],"previousBlock":"b-e7dd1dd7768c13f1ecb98165","nextBlock":"b-baccd3c4d98ebdcf709fff0c","prerequisites":[{"id":"b-a06529c46ba769c1bc0f0e50","url":"#b-a06529c46ba769c1bc0f0e50","role":"section-title"},{"id":"b-4ce3918203caf3eefd2abef9","url":"#b-4ce3918203caf3eefd2abef9","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-baccd3c4d98ebdcf709fff0c","type":"p","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-baccd3c4d98ebdcf709fff0c","content":"先找唯一蓝杯，再从其他杯子中筛选：红杯满足 $32.5<126.5$，应返回红杯框。书也在蓝杯左侧，却不满足 cup 类别约束。若蓝杯缺失、出现多个蓝杯或存在多个满足关系的目标，程序必须有明确的无目标/多义处理，不能随意返回最高颜色分的对象。","sha256":"a7e6186a3879cf00b7b4e6691c1a5fe72292766a0c2120f0667b376036a914b6","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"32.5<126.5","definitionContext":"“蓝杯左边的杯子”：目标、参照物、关系三者分开"}],"previousBlock":"b-60556010adce6f629d92884d","nextBlock":"b-7437a583e1091e603e0afafd","prerequisites":[{"id":"b-a06529c46ba769c1bc0f0e50","url":"#b-a06529c46ba769c1bc0f0e50","role":"section-title"},{"id":"b-4ce3918203caf3eefd2abef9","url":"#b-4ce3918203caf3eefd2abef9","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-7437a583e1091e603e0afafd","type":"p","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-7437a583e1091e603e0afafd","content":"这里说明的是一套显式关系语义。“完全在左侧”可以改用 $x_{max}(a)<x_{min}(r)$，与中心比较在重叠情况下不同。把语言问题落实为区域时，要先解释采用哪种语义，再谈模型是否正确。图上的坐标约定是半开边界框，不是逐像素包含端点的面积公式。","sha256":"374d2bfb115c9901efd89d4182bd0a05ca1076977308b072a18ea0d33bcb6d4a","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"x_{max}(a)<x_{min}(r)","definitionContext":"“蓝杯左边的杯子”：目标、参照物、关系三者分开"}],"previousBlock":"b-baccd3c4d98ebdcf709fff0c","nextBlock":"b-2f494a40f81c0af5249e317e","prerequisites":[{"id":"b-a06529c46ba769c1bc0f0e50","url":"#b-a06529c46ba769c1bc0f0e50","role":"section-title"},{"id":"b-4ce3918203caf3eefd2abef9","url":"#b-4ce3918203caf3eefd2abef9","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-2f494a40f81c0af5249e317e","type":"h2","scope":"grounding-target-reference-relation","url":"https://codingai-lec03.pages.dev/course/blocks/grounding-target-reference-relation.html#b-2f494a40f81c0af5249e317e","content":"## 继续阅读与实验","sha256":"2b9ccef7e1220e2250cceff1abd8bdd736b42f8da5a223f26b029f44b962e807","chapter":{"id":"III","title":"视觉与语言"},"unit":{"id":"grounding-object-contract","title":"一句话怎样落实为正确对象的区域？"},"page":{"id":"reading-opening","title":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","number":1},"sectionTitle":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","currentQuestion":"“蓝杯左边的杯子”：目标、参照物、关系三者分开","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-7437a583e1091e603e0afafd","nextBlock":null,"prerequisites":[{"id":"b-a06529c46ba769c1bc0f0e50","url":"#b-a06529c46ba769c1bc0f0e50","role":"section-title"},{"id":"b-4ce3918203caf3eefd2abef9","url":"#b-4ce3918203caf3eefd2abef9","role":"problem-context"}],"relatedBlocks":[]}]}
