从视频到可探索的知识From video to explorable knowledge
AI-EdVision 从视频处理开始，提取字幕与关键内容，再构建知识图谱和问答能力。目标不是增加一个聊天框，而是让学习者能够在观看、检索、提问、记录和复习之间自然切换。
AI-EdVision starts with video processing, extracts subtitles and key content, then builds knowledge graph and Q&A capabilities. The goal is not to add another chat box, but to let learners move naturally between watching, searching, asking, note-taking, and review.
- 视频与字幕Video and subtitles支持视频上传与链接导入，使用 Whisper 提取字幕并组织时间片段。Video upload and URL import, with Whisper transcription organized into timestamped segments.
- 知识图谱Knowledge graph从内容中整理知识点与关联关系，支持可视化浏览和学习线索发现。Extracted concepts and relationships for visual exploration and learning-path discovery.
- 智能问答AI Q&A把视频上下文、字幕与学习记录用于针对性问答，而不是脱离材料泛化回答。Grounded answers in video context, subtitles, and learning notes rather than generic responses.
- 完整工程Complete systemFastAPI · SQLAlchemy 2.0 · Neo4j · Whisper · Ollama · Vue 3 · D3.js
从问题到验证From problem to validation
我最有成就感的是把“看完教学视频后，信息仍然分散、难以继续学习”的问题，做成可探索、可提问、可复习的学习系统。先将视频、字幕和时间片段整理为知识结构，再把知识图谱与基于上下文的问答放进同一学习界面。
The work I am proudest of is turning the problem of fragmented learning after watching educational video into a system that can be explored, questioned, and reviewed. I first organized video, transcripts, and timestamps into knowledge structure, then brought knowledge-graph exploration and context-grounded Q&A into one learning interface.
从产品设计、后端迁移到 AI 能力、前端联调和竞赛材料均由我推进。结果以可核验的竞赛成果衡量：项目获第 18 届中国大学生计算机设计大赛河南省一等奖、全国三等奖。
I drove product design, backend migration, AI capabilities, frontend integration, and competition materials. The result is independently verifiable: the project won first prize in Henan and third prize nationally in the 18th Chinese Collegiate Computing Competition.
真实界面Working interface
获奖证明Competition results
项目获得第 18 届中国大学生计算机设计大赛河南省一等奖、全国三等奖。GitHub 源码当前保持闭源，页面只展示可公开的产品能力与成果。
The project received first prize in Henan and third prize nationally in the 18th Chinese Collegiate Computing Competition. The source repository remains private; only public product capabilities and results are shown here.
