{"id":51236769,"url":"https://github.com/shenliuming/inlook-yolo-model-lab","last_synced_at":"2026-06-28T21:01:34.346Z","repository":{"id":361557049,"uuid":"1254912267","full_name":"shenliuming/inlook-yolo-model-lab","owner":"shenliuming","description":"INLOOK Studio is an open-source AI content creation workbench for turning owned or authorized scripts, audio, and video materials into reusable talking-head content.","archived":false,"fork":false,"pushed_at":"2026-06-18T10:03:30.000Z","size":6906,"stargazers_count":2,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-18T11:29:01.000Z","etag":null,"topics":["ai-education","computer-vision","model-testing","object-detection","ultralytics","yolo","yolov8"],"latest_commit_sha":null,"homepage":"https://in-look.cn/yolo/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/shenliuming.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-31T06:42:52.000Z","updated_at":"2026-06-18T10:03:44.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/shenliuming/inlook-yolo-model-lab","commit_stats":null,"previous_names":["shenliuming/inlook-yolo-model-lab"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/shenliuming/inlook-yolo-model-lab","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shenliuming%2Finlook-yolo-model-lab","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shenliuming%2Finlook-yolo-model-lab/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shenliuming%2Finlook-yolo-model-lab/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shenliuming%2Finlook-yolo-model-lab/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/shenliuming","download_url":"https://codeload.github.com/shenliuming/inlook-yolo-model-lab/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shenliuming%2Finlook-yolo-model-lab/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34903523,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-28T02:00:05.809Z","response_time":54,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-education","computer-vision","model-testing","object-detection","ultralytics","yolo","yolov8"],"created_at":"2026-06-28T21:01:31.089Z","updated_at":"2026-06-28T21:01:34.328Z","avatar_url":"https://github.com/shenliuming.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# INLOOK YOLO Model Lab\n\n一个用于图片、视频和摄像头识别测试的本地 YOLO 网页实验平台。\n\nINLOOK YOLO Model Lab is a local-first YOLO web lab for image, video, and camera-based recognition tests. It is designed for computer vision learning, model validation, demo recording, and lightweight internal tooling.\n\n## Online Demo\n\n- [https://in-look.cn/yolo/](https://in-look.cn/yolo/)\n\n## Architecture Docs\n\n- [正式架构说明](docs/architecture.md)\n- [项目架构图 + 数据流图 + 目录职责说明](docs/project_architecture_and_data_flow.md)\n\n## Frontend Roles\n\n- `inlook-studio-web` 是正式产品主前端\n- `apps/yolo-web` 保留为实验室前端\n\n## Features\n\n- 图片识别\n- 视频识别\n- 摄像头识别\n- OBS 虚拟摄像头\n- 模型切换\n- 运行日志\n- JSON 测试报告\n- 结果下载\n- 字幕识别\n- TTS 配音生成\n\n## Screenshots\n\nProject screenshots can be placed here:\n\n- `docs/images/home.png`\n- `docs/images/yolo-demo.png`\n\nCurrent repository state:\n\n- Screenshot placeholders only\n- No fake images generated in README\n\n## Quick Start\n\n### 1. Install `uv`\n\n```bash\nbrew install uv\n```\n\n### 2. Install backend dependencies\n\n```bash\ncd apps/yolo-api\nuv venv --python 3.11\nsource .venv/bin/activate\nuv pip install -r requirements.txt\n```\n\n### 2.1 Prepare CosyVoice for local TTS\n\n```bash\ncd apps/yolo-api\nuv pip install -r requirements.txt\n```\n\nConfigure the CosyVoice model directory before starting the backend:\n\n```bash\nexport TTS_ENGINE=cosyvoice\nexport COSYVOICE_MODEL_DIR=pretrained_models/CosyVoice2-0.5B\nexport COSYVOICE_DEVICE=auto\nexport COSYVOICE_SAMPLE_RATE=24000\n```\n\nINLOOK Studio now routes builtin voices, custom voices, and current-video voices\nthrough CosyVoice only. MOSS-TTS is deprecated and is not used as a fallback.\n\n### 3. Start FastAPI\n\n```bash\ncd apps/yolo-api\nuv run uvicorn app:app --reload --host 127.0.0.1 --port 7860\n```\n\n### 4. Start frontend\n\n```bash\ncd apps/yolo-web\nnpm install\nnpm run dev -- --host 127.0.0.1 --port 5173\n```\n\n### 5. Open the local page\n\n- Frontend: `http://127.0.0.1:5173`\n- Backend health: `http://127.0.0.1:7860/api/health`\n\nIf you want to use the old local backend port `8000`, start Vite with:\n\n```bash\nVITE_API_TARGET=http://127.0.0.1:8000 npm run dev -- --host 127.0.0.1 --port 5173\n```\n\n## Compliance\n\n本项目仅用于计算机视觉学习、模型测试和内容创作。  \n系统只输出识别结果，不提供任何游戏控制、自动操作或绕过机制。\n\n## Want To Train Your Own Model?\n\n如果你正在训练自己的 YOLO 模型，可以先准备这些材料：\n\n- `data.yaml`\n- 训练结果图\n- 少量样本图\n- 测试视频\n\n这样更容易判断问题可能出在：\n\n- 数据本身\n- 标注质量\n- 类别设计\n- 训练参数\n- 场景差异\n\n## Project Structure\n\n```text\ninlook-yolo-model-lab/\n├── apps/\n│   ├── yolo-api/\n│   │   ├── app.py\n│   │   ├── app/\n│   │   │   ├── main.py\n│   │   │   ├── controllers/\n│   │   │   ├── services/\n│   │   │   ├── clients/\n│   │   │   ├── config/\n│   │   │   ├── common/\n│   │   │   └── utils/\n│   │   ├── models/\n│   │   ├── requirements.txt\n│   │   └── Dockerfile\n│   └── yolo-web/\n│       ├── src/api/\n│       ├── src/components/\n│       ├── src/App.vue\n│       └── Vue frontend\n├── assets/\n│   └── demo/\n├── docs/\n├── deploy/\n├── pretrained_models/\n│   └── CosyVoice2-0.5B/    # local model directory, not committed\n└── README.md\n```\n\n## AI Content Workflow\n\nThe content workflow is separate from the YOLO vision lab. It currently includes:\n\n- Material Intake\n- Subtitle Recognition\n- TTS Voice Generation\n\nSubtitle-related files are integrated into the backend service layer:\n\n- `apps/yolo-api/app/services/subtitle_tool/subtitle_pack.py`\n- `apps/yolo-api/app/services/subtitle_tool/burn_subtitles.py`\n- `apps/yolo-api/app/services/subtitle_tool/check_env.py`\n\nTTS-related files:\n\n- `apps/yolo-api/app/controllers/tts_controller.py`\n- `apps/yolo-api/app/services/tts_service.py`\n- `apps/yolo-api/app/services/tts_engines/cosyvoice_engine.py`\n\nTTS runtime output:\n\n- `apps/yolo-api/runtime/content_lab/tts/tasks/{task_id}/inputs`\n- `apps/yolo-api/runtime/content_lab/tts/tasks/{task_id}/outputs`\n- `apps/yolo-api/runtime/content_lab/tts/tasks/{task_id}/run.log`\n\nHelpful docs:\n\n- `docs/subtitle-workflow/example_usage.md`\n- `docs/subtitle-workflow/PRODUCTION_WORKFLOW.md`\n\nQuick env check:\n\n```bash\nuv run python apps/yolo-api/app/services/subtitle_tool/check_env.py\n```\n\n## Backend\n\nRecommended startup:\n\n```bash\ncd apps/yolo-api\nuv venv --python 3.11\nuv pip install -r requirements.txt\nuv run uvicorn app:app --reload --host 127.0.0.1 --port 7860\n```\n\nAPI endpoints:\n\n- `GET /api/v1/health`\n- `GET /api/v1/vision/health`\n- `GET /api/v1/vision/models`\n- `POST /api/v1/vision/models/select`\n- `POST /api/v1/vision/images/detect`\n- `POST /api/v1/vision/videos/detect`\n- `POST /api/v1/vision/realtime/detect`\n- `GET /api/v1/vision/tasks/{task_id}`\n- `GET /api/v1/vision/tasks/{task_id}/files/{filename}`\n- `GET /api/v1/content-lab/health`\n- `GET /api/v1/content-lab/materials/health`\n- `POST /api/v1/content-lab/materials/tasks`\n- `GET /api/v1/content-lab/materials/tasks/{task_id}`\n- `GET /api/v1/content-lab/materials/tasks/{task_id}/files/{filename}`\n- `GET /api/v1/content-lab/subtitles/health`\n- `POST /api/v1/content-lab/subtitles/tasks`\n- `GET /api/v1/content-lab/subtitles/tasks/{task_id}`\n- `POST /api/v1/content-lab/subtitles/tasks/{task_id}/reburn`\n- `GET /api/v1/content-lab/subtitles/tasks/{task_id}/files/{filename}`\n- `GET /api/v1/content-lab/tts/health`\n- `POST /api/v1/content-lab/tts/tasks`\n- `GET /api/v1/content-lab/tts/tasks/{task_id}`\n- `GET /api/v1/content-lab/tts/tasks/{task_id}/files/{filename}`\n\nCompatibility endpoints are still available:\n\n- `GET /api/models`\n- `POST /api/detect/image`\n- `POST /api/detect/video`\n- `POST /api/realtime/detect`\n- `POST /api/materials/tasks`\n- `GET /api/materials/tasks/{task_id}`\n- `GET /api/materials/tasks/{task_id}/files/{filename}`\n\nStatic result paths:\n\n- `/outputs`\n- `/reports`\n\n## Frontend\n\n```bash\ncd apps/yolo-web\nnpm install\nnpm run dev -- --host 127.0.0.1 --port 5173\n```\n\nOpen:\n\n- `http://127.0.0.1:5173`\n\nFrontend routes:\n\n- `/`\n- `/vision-lab`\n- `/vision-lab/model-test`\n- `/vision-lab/image`\n- `/vision-lab/video`\n- `/vision-lab/realtime`\n- `/content-workflow`\n- `/content-workflow/material-intake`\n- `/content-workflow/subtitle-recognition`\n- `/content-workflow/tts`\n- `/content-lab`\n- `/content-lab/material-intake`\n- `/content-lab/subtitle-recognition`\n- `/content-lab/tts`\n\nLegacy redirects:\n\n- `/material-intake` -\u003e `/content-workflow/material-intake`\n- `/content-intake` -\u003e `/content-workflow/material-intake`\n- `/vision-lab/image-detect` -\u003e `/vision-lab/image`\n- `/vision-lab/video-detect` -\u003e `/vision-lab/video`\n- `/vision-lab/realtime-detect` -\u003e `/vision-lab/realtime`\n\nShared frontend modules:\n\n- `src/api/client.js`\n- `src/api/vision.js`\n- `src/api/workflow.js`\n- `src/api/contentLabApi.js`\n- `src/components/StatusCard.vue`\n- `src/components/TaskLog.vue`\n- `src/components/FileDownloadList.vue`\n\n## Camera Mode\n\n- Frontend uses `getUserMedia()` to open camera devices\n- Supports standard cameras and OBS virtual camera selection\n- Captures frames and sends them to the backend for YOLO inference\n- Draws boxes, labels, and confidence values on overlay canvas\n- Shows the latest JSON result in realtime mode\n\n## Models\n\nDefault custom model:\n\n- `apps/yolo-api/models/inlook/best.pt`\n\nScanned model directories:\n\n- `apps/yolo-api/models/official/*.pt`\n- `apps/yolo-api/models/inlook/*.pt`\n\nSecurity-related constraints:\n\n- Model files stay in backend-only storage\n- Nginx does not expose model directories\n- FastAPI does not mount model directories as static paths\n- `/api/models` returns display metadata instead of real model file paths\n- Report JSON does not expose real model file paths\n- Image upload limit: `10MB`\n- Video upload limit: `200MB`\n- Realtime frame upload limit: `4MB`\n- Allowed image types: `jpg/jpeg/png`\n- Allowed video type: `mp4`\n- Basic IP rate limiting is enabled in backend\n- Old uploads / outputs / reports are cleaned periodically\n- Optional `INLOOK_API_KEY` is supported for internal access control\n\n## Internal API Key\n\nIf you want to restrict who can call the backend:\n\n```bash\nINLOOK_API_KEY=your-secret-key\n```\n\nIf frontend should attach the same key automatically before build:\n\n```bash\nVITE_INTERNAL_API_KEY=your-secret-key\n```\n\nThe frontend will send:\n\n```txt\nX-INLOOK-Key: your-secret-key\n```\n\n## Docker Deployment\n\nThe project already includes a Docker-based deployment structure for ECS / VM deployment.\n\nFiles already included:\n\n- `apps/yolo-api/Dockerfile`\n- `apps/yolo-web/Dockerfile`\n- `deploy/nginx.conf`\n- `docker-compose.yml`\n- `.dockerignore`\n\n### Before deployment\n\nMake sure the server already has:\n\n- Docker\n- Docker Compose Plugin\n- Model file `apps/yolo-api/models/inlook/best.pt`\n\nOptional official models:\n\n- `apps/yolo-api/models/official/yolo11n.pt`\n- `apps/yolo-api/models/official/yolo11s.pt`\n- `apps/yolo-api/models/official/yolov8n.pt`\n\n### Start\n\n```bash\ndocker compose up -d --build\n```\n\nAfter startup:\n\n- Frontend: `http://your-server-ip/`\n- Backend health: `http://your-server-ip/api/health`\n\n### Stop\n\n```bash\ndocker compose down\n```\n\n### Docker cache note\n\nFor normal backend updates:\n\n```bash\ndocker compose build backend\ndocker compose up -d backend\n```\n\nAvoid using:\n\n```bash\ndocker compose build --no-cache backend\n```\n\nbecause it forces large dependencies such as `torch`, `ultralytics`, and `opencv` to be downloaded again.\n\n## Deploy Under `in-look.cn/yolo/`\n\nIf you want to mount the frontend under an existing website path:\n\n- Page entry: `https://in-look.cn/yolo/`\n- Backend API: `https://in-look.cn/yolo/api/*`\n- Result files: `https://in-look.cn/yolo/outputs/*`\n- Report files: `https://in-look.cn/yolo/reports/*`\n\nBuild frontend with subpath base:\n\n```bash\ncd apps/yolo-web\nVITE_PUBLIC_BASE=/yolo/ npm run build\n```\n\nThen publish `apps/yolo-web/dist/` to:\n\n```bash\n/var/www/in-look.cn/html/yolo/\n```\n\n## Notes\n\nPlease do not commit:\n\n- model files\n- videos\n- audio files\n- training datasets\n- generated subtitle outputs\n- generated recognition outputs\n\n## INLOOK Studio LLM 配置\n\n提示词改写、文案校对和标题生成统一走 OpenAI-compatible Chat Completions 服务。不要把真实 API Key 写进代码，使用环境变量：\n\n```bash\nLLM_PROVIDER=openai_compatible\nLLM_BASE_URL=https://your-model-service/v1\nLLM_API_KEY=your_api_key\nLLM_MODEL=your-model-name\nLLM_TIMEOUT_SECONDS=60\n```\n\n未配置时 `GET /api/v1/ai/status` 会返回 `available=false`，前端会禁用 AI 改写按钮，不会生成 mock 文案。\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fshenliuming%2Finlook-yolo-model-lab","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fshenliuming%2Finlook-yolo-model-lab","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fshenliuming%2Finlook-yolo-model-lab/lists"}