发布前检查清单

适用于网站首次公开、数据版本更新和算法资源接入。当前交付是本地网站框架；数据标准化与 AI 算法已由项目方完成，下面检查的是公开材料、网站映射及发布条件。没有证据的项目保持“待核实”，不要勾选。

网页版清单为只读参考，不保存审核状态；请在项目自己的发布记录中保存检查结果、证据与负责人。

1. 公开范围与身份

项目负责人批准本次公开范围、目标仓库和站点。
正式名称、所属机构、作者顺序、联系人及其公开授权已核实。
页面清楚区分真实发布内容、演示样例、合成示意图和待补充信息。
DEMO 不计入真实样本量、类别占比、数据规模、算法性能或研究结论。
用于演示的条目已移除，或继续明确标注并与真实数据区分。

2. 数据与索引映射

网站索引与现有标准的字段对应关系、标准版本及解释经负责人确认。
样本 ID 唯一、稳定；标题、类别、描述与真实资料一致。
维度顺序、坐标约定、单位、体素间距、波长及缺失值语义已核实。
阵列/采集/重建来源及参数能够追溯；不能公开的部分留在受控记录中。
数据划分是实际采用的划分；同一对象、实验或关联样本跨集合泄漏风险已检查。
元数据通过 Schema 校验；必要扩展同步修改 Schema、展示逻辑与说明。
索引统计与已获准公开的实际文件一致。

3. 伦理、隐私与去标识化

涉及人体、动物或其他受限制资料时，项目方已核实相应审批、同意和共享条件。
发布内容符合原始同意和授权范围；不能因为去标识化而默认允许公开。
文件名、文件头、嵌入元数据、缩略图、图中标注和压缩包内容均已检查。
已移除身份标识、内部编号关联表、敏感日期、内部存储路径和凭据。
可识别解剖特征、稀有属性组合及重新识别风险已由具备资格的负责人评估。
公开 JSON、代码仓库历史和站点产物均不含受限信息。

此清单是发布操作提醒，不替代机构审批、数据治理或专业法律意见。

4. 许可、作者与引用

数据、算法代码、权重、网站代码和第三方素材分别有明确且相容的使用条件。
作者、贡献、资助和致谢经相关人员确认。
已核验引用文本、版本、发布日期和永久标识符；没有 DOI 时不虚构。
引用指向实际发布版本，更新内容保留版本记录和变更摘要。
必要的适用范围、研究用途说明、限制及撤回联系方式清楚可见。

5. 数据托管与下载

数据文件托管位置经批准；每个下载链接都已实际访问并与描述一致。
文件格式、压缩方式、大小和 SHA-256 校验值从最终文件计算。
用空白会话检查公开资源是否可用；受控资源明确说明申请或登录条件。
没有泄露临时签名 URL、认证令牌或仅供内部使用的分享链接。
对跨域获取、断点续传和大文件下载需求进行实际验证，不把浏览器行为当作保证。
已核对托管服务容量、流量、保留期限及费用；GitHub Pages 仅承载轻量站点与索引。
文件更新后同步版本和校验值；失效资源不会继续显示为可下载。

此清单记录的 GitHub Pages 已发布站点上限为 1 GB，月带宽软限制为 100 GB；实际发布前再次查看官方限制。普通 Git 文件也受平台限制，大型扫描与权重应另行规划托管。

6. 已有算法的网站接入

算法名称、用途、输入输出和适用范围与现有成果一致。
公开代码、模型、说明书及论文链接的实际可用性分别标注。
提供对应提交/版本、运行环境、依赖、硬件需求、随机种子和必要配置。
在获准提供的真实样例上验证推理命令和输出，记录预处理及后处理。
指标附数据划分、定义、单位、实验条件及可追溯结果；不使用占位数字。
明确网站是否只有资源介绍；只有实际实现后才能声明在线推理或数据处理功能。

详见算法接入指南。

7. 页面质量与可访问性

桌面、平板和窄屏手机均无重要内容溢出或遮挡。
筛选、搜索、清空、空结果、详情开关和重复操作均可使用。
键盘可到达全部交互；焦点可见，弹窗能关闭并正确恢复焦点。
标签、替代文本、颜色对比、状态提示和减少动态效果得到检查。
JSON 加载失败有说明；缺失字段、空值与长标题不会导致脚本报错。
使用 HTTP 本地预览并检查浏览器控制台和网络错误。
所有站内路径支持 GitHub Pages 仓库子路径，不能只在域名根路径通过。
中 / EN 切换可更新完整正文、导航、无障碍标签与已打开的详情；存储可用时可记住选择。
两种语言均保留合成 DEMO 警示，并明确尚未提供或核验的研究材料。

8. 最终发布和验证

检查 .github/workflows/pages.yml 的 _site 文件白名单。
_site 不含测试报告、浏览器截图、源码 ZIP、私密数据或未审核文件。
已审阅本次提交，确认仅在授权后手动点击 Run workflow。
记录实际部署提交、工作流结果和站点地址。
发布后复查首页、指南、JSON、外部下载及引用；记录检查日期。
指定维护人、问题反馈路径和必要时撤下内容的操作责任人。

未完成的检查应列出负责人及需要的材料。若决定先发布框架预览，应保留明显的 DEMO 与待接入标记，不能宣传为真实数据集已开放下载。

Pre-release checklist

Use this checklist for the first public website release, data-version updates, and algorithm-resource integration. The current deliverable is a local website framework. The project team has already completed data standardization and the AI algorithm. These checks cover public materials, website mapping, and release conditions. Leave unsupported items as “Pending verification”; do not check them off.

The website checklist is a read-only reference and does not save review status. Keep findings, evidence, and responsible people in the project's own release record.

1. Public scope and identity

The project lead has approved this release's public scope, target repository, and website.
The official name, affiliation, author order, contacts, and their permission to be publicly named have been verified.
Pages clearly distinguish actual release content, demo samples, synthetic illustrations, and information still to be supplied.
DEMO records do not count toward real sample counts, category proportions, data scale, algorithm performance, or research conclusions.
Demo records have been removed or remain clearly labeled and separated from real data.

2. Data and index mapping

The responsible lead has confirmed the field mapping between the website index and existing standard, the standard version, and field interpretations.
Sample IDs are unique and stable; titles, categories, and descriptions match actual materials.
Dimension order, coordinate conventions, units, voxel spacing, wavelength, and missing-value meanings have been verified.
Array/acquisition/reconstruction sources and parameters are traceable; nonpublic information remains in controlled records.
Dataset splits match those actually used, and leakage risks across sets involving the same subject, experiment, or related samples have been checked.
Metadata passes Schema validation. Any required extensions are reflected in the Schema, rendering logic, and documentation.
Index statistics match the actual files approved for public release.

3. Ethics, privacy, and de-identification

For human, animal, or other restricted materials, the project team has verified the necessary approvals, consent, and sharing conditions.
Released content stays within the original consent and authorization. De-identification alone is not assumed to permit publication.
Filenames, file headers, embedded metadata, thumbnails, image annotations, and archive contents have all been checked.
Identifiers, internal-ID linkage tables, sensitive dates, internal storage paths, and credentials have been removed.
A qualified responsible person has assessed identifiable anatomical features, rare combinations of attributes, and re-identification risk.
Public JSON, repository history, and website build artifacts contain no restricted information.

This checklist is an operational release reminder, not a substitute for institutional approval, data governance, or professional legal advice.

4. Licenses, authorship, and citations

Data, algorithm code, weights, website code, and third-party assets each have clear and compatible terms of use.
Relevant people have confirmed authorship, contributions, funding, and acknowledgments.
Citation text, versions, release dates, and persistent identifiers have been verified. No DOI is invented when none exists.
Citations identify the actual released version; updates retain version history and a change summary.
Intended scope, research-use statements, limitations, and withdrawal contacts are clearly visible where needed.

5. Data hosting and downloads

File hosting is approved, and every download link has been opened and checked against its description.
File format, compression, size, and SHA-256 checksums are determined from the final files.
Public resources have been checked in a fresh session; controlled resources clearly state application or sign-in requirements.
No temporary signed URLs, authentication tokens, or internal-only sharing links are exposed.
Cross-origin fetching, resumable downloads, and large-file requirements have been tested where relevant. Browser behavior is not assumed to be guaranteed.
Hosting capacity, bandwidth, retention, and costs have been checked. GitHub Pages serves only the lightweight website and index.
Versions and checksums are updated after file changes; unavailable resources no longer appear downloadable.

The limits recorded in this guide are 1 GB for a published GitHub Pages site and a soft monthly bandwidth limit of 100 GB. Check the official limits again before deployment. Ordinary Git files also have platform limits; plan separate hosting for large scans and weights.

6. Website integration of the existing algorithm

The algorithm name, purpose, inputs, outputs, and scope match the existing work.
Actual availability is labeled separately for public code, models, manuals, and papers.
The corresponding commit/version, environment, dependencies, hardware requirements, random seeds, and necessary configuration are provided.
Inference commands and outputs have been verified on a real example approved for sharing, with preprocessing and postprocessing recorded.
Metrics include dataset splits, definitions, units, experimental conditions, and traceable results. No placeholder numbers are used.
The site clearly states whether it only presents resources. Online inference or data-processing features are claimed only after implementation.

See the algorithm integration guide.

7. Page quality and accessibility

Important content is neither clipped nor obscured on desktop, tablet, or narrow mobile screens.
Filtering, search, reset, empty results, detail opening/closing, and repeated operations work.
All interactions are keyboard-accessible; focus is visible, dialogs close, and focus returns correctly.
Labels, alternative text, color contrast, status messages, and reduced-motion behavior have been checked.
JSON loading errors are explained; missing fields, nulls, and long titles do not cause script errors.
The site has been previewed over local HTTP, with browser console and network errors checked.
All internal paths work under a GitHub Pages repository subpath, not only at the domain root.
The 中 / EN switch updates complete page content, navigation, accessibility labels, and open detail views; the selected language persists when storage is available.
Both languages preserve synthetic DEMO warnings and clearly mark research materials that have not yet been provided or verified.

8. Final release and verification

The _site allowlist in .github/workflows/pages.yml has been checked.
_site excludes test reports, browser screenshots, source ZIPs, private data, and unreviewed files.
The changes have been reviewed, and Run workflow is clicked manually only after authorization.
The actual deployed commit, workflow result, and website URL are recorded.
After release, the homepage, guides, JSON, external downloads, and citations are checked again, with the check date recorded.
A maintainer, feedback route, and person responsible for withdrawing content if needed have been designated.

For unfinished checks, identify the responsible person and required materials. If releasing a framework preview first, retain prominent DEMO and pending-integration labels. Do not advertise it as a real dataset already open for download.