{"id":47788102,"url":"https://github.com/sairam-s0/activelabelingsystem","last_synced_at":"2026-04-03T15:05:43.185Z","repository":{"id":331089417,"uuid":"1125206830","full_name":"sairam-s0/ActiveLabelingSystem","owner":"sairam-s0","description":"LabelOps – A local-first AI-assisted data labeling system with continual learning and background model training.","archived":false,"fork":false,"pushed_at":"2026-04-02T10:44:47.000Z","size":54773,"stargazers_count":6,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-04-03T00:56:09.515Z","etag":null,"topics":["ai","computer-vision","data-engineering","labeling-tool","learners-implementation","local","machine-learning"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/sairam-s0.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","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":"2025-12-30T10:19:11.000Z","updated_at":"2026-04-02T10:44:51.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/sairam-s0/ActiveLabelingSystem","commit_stats":null,"previous_names":["sairam-s0/labelops","sairam-s0/activelabelingsystem"],"tags_count":2,"template":false,"template_full_name":null,"purl":"pkg:github/sairam-s0/ActiveLabelingSystem","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sairam-s0%2FActiveLabelingSystem","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sairam-s0%2FActiveLabelingSystem/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sairam-s0%2FActiveLabelingSystem/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sairam-s0%2FActiveLabelingSystem/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sairam-s0","download_url":"https://codeload.github.com/sairam-s0/ActiveLabelingSystem/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sairam-s0%2FActiveLabelingSystem/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31359124,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-03T14:24:47.623Z","status":"ssl_error","status_checked_at":"2026-04-03T14:24:32.380Z","response_time":107,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"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","computer-vision","data-engineering","labeling-tool","learners-implementation","local","machine-learning"],"created_at":"2026-04-03T15:05:42.428Z","updated_at":"2026-04-03T15:05:43.158Z","avatar_url":"https://github.com/sairam-s0.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# ActiveLabelingSystem\n\nFormerly known as **LabelOps**.\n\nLocal-first AI-assisted image labeling with active learning, manual box tools, background retraining, and dataset version snapshots.\n\n## What is new\n\n- Modular app structure (`src/app`, `src/core`, `src/features`) for cleaner maintenance.\n- Active learning image prioritization when loading folders (unlabeled images first, sorted by uncertainty).\n- Retraining policy engine with multi-signal checks (sample count, time, entropy shift, class balance, confidence drift).\n- Dataset versioning from UI with metadata, hash integrity, and manifest tracking.\n- Label format selection at folder load time: `COCO JSON` or `Plain JSON`.\n- Safer persistence:\n  - autosave per folder (`labels_autosave.json`)\n  - internal state store (`.labels_internal.json`)\n  - atomic file writes for exports.\n- Improved manual labeling UX:\n  - floating toolbox\n  - quick class switching (1-9)\n  - undo/delete shortcuts\n  - inline floating action toolbar.\n- Better training controls in UI:\n  - force retrain\n  - live queue/training status\n  - shadow model promotion with validation warning flow.\n\n## Features\n\n### Active learning and triage\n\n- Entropy-aware detection metadata is attached to predictions.\n- Folder loading prioritizes unlabeled images for high-value review.\n- Replay buffer preserves historical samples for continual learning.\n\n### Retraining workflow\n\n- Background training via Ray shadow trainer.\n- Training trigger policy requires minimum sample count plus at least one urgency signal.\n- Promotion flow supports validation and explicit override confirmation.\n\n### Dataset versioning\n\n- Create version snapshots from current labels.\n- Stores YOLO-style labels, copied images, metadata, and a dataset hash.\n- Keeps lineage and latest pointer in `src/datasets/manifest.json`.\n\n### Label export\n\n- `Plain JSON` output (`labels.json`) for direct app consumption.\n- `COCO JSON` output (`labels_coco.json`) for downstream ML pipelines.\n\n### Manual labeling mode\n\n- Draw boxes directly on the canvas.\n- Per-box class assignment with deterministic class colors.\n- Save-and-next loop without leaving manual mode.\n\n## Project structure\n\n```text\nActiveLabelingSystem/\n  images/\n  src/\n    app/\n      window.py\n      dialogs.py\n      state.py\n      actions.py\n    core/\n      data_manager.py\n      entropy.py\n      sample_selector.py\n      retrain_policy.py\n      dataset_versioner.py\n      replay_buffer.py\n      shadow_trainer.py\n      training_orchestrator.py\n      model_manager.py\n      feedback_validator.py\n    features/\n      manual.py\n      shortcut_manager.py\n      shortcut_config.py\n      toolbar_manager.py\n      toolbar_widget.py\n      toolbar_styles.py\n    datasets/\n    models/\n    main.py\n    requirements.txt\n    run_tests.bat\n  README.md\n  GUID.md\n```\n\n## Installation\n\n```bash\ngit clone https://github.com/sairam-s0/ActiveLabelingSystem.git\ncd ActiveLabelingSystem\npython -m venv .venv\n# Windows\n.venv\\Scripts\\activate\n# Linux/macOS\n# source .venv/bin/activate\n\npip install .\n```\n\nOptional: install CUDA-enabled PyTorch for GPU inference/training from https://pytorch.org/get-started/locally/.\nUsers can install it directly with:\n\n```bash\npip install Active-Labeling-System\n```\n\n## Run the app\n\nAfter install:\n\n```bash\nals\n```\n\n## Publish to PyPI (GitHub Actions)\n\nThis repo includes `.github/workflows/publish-pypi.yml`.\n\n1. In PyPI, configure Trusted Publisher for this GitHub repo/workflow:\n   - Owner: `sairam-s0`\n   - Repository: `ActiveLabelingSystem`\n   - Workflow: `publish-pypi.yml`\n2. Bump version in `pyproject.toml`.\n3. Push a tag like `v0.1.1`.\n4. GitHub Actions builds and publishes automatically to PyPI.\n## How to use\n\n### 1. Select image folder and output format\n\n- Click `Select Folder`.\n- Choose output format:\n  - `COCO JSON` -\u003e writes `labels_coco.json`\n  - `Plain JSON` -\u003e writes `labels.json`\n- The app also keeps internal state in `.labels_internal.json` and autosave in `labels_autosave.json` inside the selected folder.\n\n![Folder Selection](./images/folder_selection.png)\n![Format Selection](./images/format_selection.png)\n\n### 2. Select classes\n\n- Click `Select Classes`.\n- Pick one or more classes.\n- Add custom classes from the same dialog when needed.\n\n![Class Selection](./images/class_selection.png)\n\n### 3. Start labeling\n\n- Click `START`.\n- Review detections and use bottom actions:\n  - `Accept (A)`\n  - `Reject (R)`\n  - `Skip (N)`\n  - `Manual (M)`\n\n![Active Learning Options](./images/active_learning_options.png)\n\n### 4. Manual mode (box drawing)\n\n![Normal GUI](./images/normal_full_gui.png)\n\n- Draw boxes by click-drag on canvas.\n- Save boxes and move next with:\n  - `Space` or `Enter` -\u003e save and next\n  - `Esc` -\u003e exit manual mode\n  - `Ctrl+Z` -\u003e undo last box\n  - `Delete` -\u003e delete last box\n  - `1..9` -\u003e switch class index\n\n![Manual Labelling](./images/manual_labelling.png)\n\n### 5. Monitor active learning and training\n\n- Left panel shows:\n  - entropy of current image\n  - queue size\n  - training progress/status.\n- Use `Force Retrain` if you want to bypass normal policy checks (still requires minimum samples).\n\n![Dataset Statistics](./images/dataset_stats.png)\n\n### 6. Version and promote\n\n- `Create Version` creates a dataset snapshot in `src/datasets/v_YYYYMMDD_HHMMSS/`.\n- `List Versions` shows stored versions and metadata.\n- `Promote Shadow` promotes trained candidate model to active model.\n\n## Output files and artifacts\n\nFor each selected image folder:\n\n- `labels.json` or `labels_coco.json` (selected format)\n- `.labels_internal.json` (internal metadata store)\n- `labels_autosave.json` (session recovery)\n\n## Notes\n\n- If Ray is unavailable, labeling still works; background training features are reduced.\n- If class mapping is not available yet, trainer creation waits until first labels are saved.\n- Restart app after model promotion for a clean reload of active weights.\n\n## Contributing\n\nPlease read `CONTRIBUTING.md`.\n\n## License\n\nMIT. See `LICENSE`.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsairam-s0%2Factivelabelingsystem","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsairam-s0%2Factivelabelingsystem","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsairam-s0%2Factivelabelingsystem/lists"}