[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80961":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":12,"contributorsCount":12,"subscribersCount":12,"size":12,"stars1d":12,"stars7d":12,"stars30d":14,"stars90d":12,"forks30d":12,"starsTrendScore":12,"compositeScore":15,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":16,"fork":16,"defaultBranch":17,"hasWiki":18,"hasPages":16,"topics":19,"createdAt":9,"pushedAt":9,"updatedAt":20,"readmeContent":21,"aiSummary":22,"trendingCount":12,"starSnapshotCount":12,"syncStatus":23,"lastSyncTime":24,"discoverSource":25},80961,"IMU-Surgical-Intention-Perception","MarkMaMed\u002FIMU-Surgical-Intention-Perception","MarkMaMed","This project is based on the surgical assistant upper limb exoskeleton with dual modes (tracking\u002Fpose maintenance) for surgical operations, which was developed by the Innovation Project of Peking University Health Science Center in 2025-2026. It is designed as an operator's intention perception model. ",null,"Python",32,0,31,1,37.1,false,"main",true,[],"2026-06-12 04:01:31","# 方案一（IMU 手术意态感知）\n\n本仓库现已清理视觉识别意图相关模块，当前仅保留 IMU 多数据集意态感知流程（JIGSAWS、Opportunity、NinaPro、PAMAP2）。\n\n## 目录\n\n```text\nconfig\u002F\n  imu_multidataset.toml\nimu_intent\u002F\n  loaders.py\n  train_multidataset.py\n  infer_from_csv.py\n  stream_demo.py\n  verify_jigsaws_layout.py\n  fetch_jigsaws_official.py\neval\u002F\n  plot_imu_dashboard.py\n  plot_imu_timeline.py\n  plot_intent_transition_graph.py\n  build_visual_gallery.py\ndocs\u002F\n  IMU意态感知程序说明.md\n  IMU可视化展示方案.md\n  JIGSAWS官方下载接入说明.md\n```\n\n## 安装\n\n```bash\npython3 -m pip install -r requirements.txt\n```\n\n## 训练\n\n```bash\npython3 -m imu_intent.train_multidataset \\\n  --config config\u002Fimu_multidataset.toml \\\n  --model-out models\u002Fimu_intent_multidataset_real.joblib \\\n  --metrics-out logs\u002Fimu_intent_metrics_real.json\n```\n\n## 推理\n\n```bash\npython3 -m imu_intent.infer_from_csv \\\n  --model models\u002Fimu_intent_multidataset_real.joblib \\\n  --input data\u002Fpublic_imu\u002FPAMAP2_Dataset\u002FProtocol\u002Fsubject101_small.csv \\\n  --output logs\u002Fimu_intent_subject101_small_predictions.csv \\\n  --summary logs\u002Fimu_intent_subject101_small_summary.json\n```\n\n## 可视化\n\n```bash\npython3 -m eval.plot_imu_dashboard --metrics logs\u002Fimu_intent_metrics_real.json --output plots\u002Fimu_dashboard.png\npython3 -m eval.plot_imu_timeline --predictions logs\u002Fimu_intent_subject101_small_predictions.csv --output plots\u002Fimu_timeline.png\npython3 -m eval.plot_intent_transition_graph --predictions logs\u002Fimu_intent_subject101_small_predictions.csv --output plots\u002Fimu_transition_graph.png\npython3 -m eval.build_visual_gallery --plots-dir plots --output plots\u002Fgallery.html\n```\n\n## JIGSAWS 手术意态感知（锁定\u002F不锁定）\n\n增强版说明：\n- 使用 151 个可解释运动学特征（速度\u002F加速度\u002Fjerk 分布、双手协同、时间形态与能量熵）\n- 引入手术任务上下文（`Knot_Tying` \u002F `Needle_Passing` \u002F `Suturing`）\n- 自动比较 `RandomForest`、`ExtraTrees`、`XGBoost`，并做分组交叉验证下的取交集\n- 在不看测试集的前提下，仅用训练集做术者分组 OOF 选择，再用全数据重拟合部署模型\n- 当前严格 OOF 结果：Accuracy `0.8417`，Macro-F1 `0.8266`，ROC-AUC `0.8958`\n- 当前自动选中方案：`XGBoost + ExtraTrees` 贪心集成，叠加因果 `EMA(alpha=0.75)` 时序平滑\n\n```bash\npython3 -m imu_intent.jigsaws_intent_program \\\n  --config config\u002Fjigsaws_intent.toml \\\n  --output-dir logs\u002Fjigsaws_intent \\\n  --model-out models\u002Fjigsaws_intent_model.joblib\n\npython3 -m eval.plot_jigsaws_intent_report \\\n  --metrics logs\u002Fjigsaws_intent\u002Fmetrics.json \\\n  --predictions logs\u002Fjigsaws_intent\u002Fwindow_predictions.csv \\\n  --output plots\u002Fjigsaws_intent_report.png\n\npython3 -m eval.plot_jigsaws_showcase \\\n  --metrics logs\u002Fjigsaws_intent\u002Fmetrics.json \\\n  --predictions logs\u002Fjigsaws_intent\u002Fwindow_predictions.csv \\\n  --output-dir plots\u002Fjigsaws_showcase\n\npython3 -m eval.plot_jigsaws_roc \\\n  --predictions logs\u002Fjigsaws_intent\u002Fwindow_predictions.csv \\\n  --metrics logs\u002Fjigsaws_intent\u002Fmetrics.json \\\n  --output-dir plots\u002Fjigsaws_roc\n```\n\n详细说明见：\n- `docs\u002FJIGSAWS手术意态感知程序.md`\n\n## JIGSAWS 原始 kinematics 接入\n\n官方入口：\n- [https:\u002F\u002Fwww.cs.jhu.edu\u002F~los\u002Fjigsaws\u002Finfo.php](https:\u002F\u002Fwww.cs.jhu.edu\u002F~los\u002Fjigsaws\u002Finfo.php)\n\n收到官方邮件下载链接后：\n\n```bash\npython3 -m imu_intent.fetch_jigsaws_official \\\n  --suturing-url \"\u003C官方邮件链接1>\" \\\n  --knot-url \"\u003C官方邮件链接2>\" \\\n  --needle-url \"\u003C官方邮件链接3>\"\n\npython3 -m imu_intent.verify_jigsaws_layout \\\n  --config config\u002Fimu_multidataset.toml \\\n  --output logs\u002Fjigsaws_layout_check.json\n```\n","该项目基于北京大学医学部2025-2026年度创新项目开发，旨在为外科手术辅助上肢外骨骼提供双模式（跟踪\u002F姿态保持）的术者意图感知模型。核心功能包括使用IMU数据集进行多模式训练、推理及可视化展示，并特别针对JIGSAWS手术任务上下文进行了优化，通过自动比较多种机器学习算法并采用分组交叉验证策略来提高模型性能。项目适合需要精确捕捉和理解外科医生操作意图的应用场景，如手术机器人辅助系统或智能医疗设备的研发。",2,"2026-06-11 04:03:00","CREATED_QUERY"]