OpenAxiom is a local PySide6 UI Lab annotation MVP and batch-safe YOLO label tool.
- Features
- Quick Start
- Windows Installation
- Restore After Windows Reinstall
- Data Folder Setup
- Backup and Recovery Policy
- Safety Notes
- Development Status
- License
- Open project with image + label directory
- Canvas zoom, pan, fit-to-window
- Bounding box selection, add, delete, category edit
- Undo / Redo (Ctrl+Z / Ctrl+Y)
- Dirty state protection
- Keyboard shortcuts: Delete / Esc / Enter / A / D
- Pre-save check (PASS / WARN / BLOCK)
- YOLO preview (dry-run)
- Single-label safe save (auto-backup + confirm + verify)
- Single-label safe restore (auto-backup + confirm + verify)
- Backup to label_backups_batch directory
- MVP total check
- Project scan (full dataset label consistency)
- Batch pre-save check dry-run
- Batch YOLO dry-run
- Zero-byte label scan
- Full-save gate check (26 items)
- Batch save plan (PASS / WARN / BLOCK)
- Batch backup plan
- Batch console: batch size 5 / 10 / 20
- Multi-batch executor (up to 20 batches, max 400 labels)
- Per-batch audit trail
- Batch restore preview
- Category filter + low-confidence review queue
- Next / previous low-confidence box
- Confirm and next
- Cross-image issue navigation
- Session change summary
git clone https://github.com/a740022938/OpenAxiom.git
cd OpenAxiom
python -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt
python lab_launch_v0.3.2.pyThe UI starts and you can open a YOLO dataset.
- Windows 10 / 11 (64-bit)
- Python 3.11 (recommended)
- Git
# 1. Clone the repository
git clone https://github.com/a740022938/OpenAxiom.git
cd OpenAxiom
# 2. Create virtual environment
python -m venv .venv
# 3. Activate virtual environment
.\.venv\Scripts\activate
# 4. Install dependencies
pip install -r requirements.txt
# 5. Launch
python lab_launch_v0.3.2.pyNote: You do NOT need to copy .venv from another machine. Always recreate it with pip install -r requirements.txt.
| Item | What to back up | Why |
|---|---|---|
| Dataset | YOUR_DATASET_ROOT | Contains images + labels |
| label_backups_batch | Full directory | Batch backup history |
| OpenAxiom_batch_audit | Full directory | Batch audit trails |
| Important reports | Manual copy | Reports and governance docs |
Do NOT back up .venv or pycache.
# 1. Install Python 3.11
# Download from https://www.python.org/downloads/
# 2. Clone the code
git clone https://github.com/a740022938/OpenAxiom.git
cd OpenAxiom
# 3. Recreate virtual environment
python -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt
# 4. Restore dataset
# Copy your dataset back to a local directory.
# Example: YOUR_DATASET_ROOT/dataset/
# 5. (Optional) Restore backups and audit
# YOUR_BACKUP_DIR/label_backups_batch/
# YOUR_AUDIT_DIR/OpenAxiom_batch_audit/
# 6. Launch and verify
python lab_launch_v0.3.2.pypython --version
pip list
python lab_launch_v0.3.2.pyYOUR_PROJECT_ROOT/ cloned from GitHub
YOUR_DATASET_ROOT/dataset/ your images + labels
YOUR_BACKUP_DIR/label_backups_batch/ batch backups
YOUR_AUDIT_DIR/OpenAxiom_batch_audit/ batch audit
These are examples, not hardcoded requirements.
OpenAxiom lets you choose your dataset at runtime.
- Click Open Project.
- Select the dataset root directory.
- The tool detects images/ and labels/ subdirectories.
- If auto-detection fails, configure paths manually.
| Layer | Content |
|---|---|
| source_only | Source code, config, scripts |
| full_backup | Source + .venv |
| data_backup | label_backups_batch + audit |
.venv contains platform-specific binaries. Always recreate with:
python -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt# Dry-run
powershell -ExecutionPolicy Bypass -File .\scripts\backup_openaxiom_source_only.ps1
# Execute
powershell -ExecutionPolicy Bypass -File .\scripts\backup_openaxiom_source_only.ps1 -Execute- Always run batch pre-save check dry-run before any batch save.
- Always run zero-byte label scan after batch save.
- If a batch fails, stop immediately.
- Test with 1-2 batches before full-dataset operations.
- All batch saves require manual confirmation per batch.
- v1.0.4 — Documentation final fix release.
- Platform: Windows 10/11, Python 3.11, PySide6.
This project is licensed under the MIT License. See LICENSE.