How modelgate-standard/modelgate works
ModelGate is a local computer-vision dataset auditor and MGS reference implementation. A user passes a ZIP or class-folder dataset to the Python API or CLI; readers normalize it into a format-independent Manifest, normative checkers evaluate structure, integrity, duplicates and balance, and an informative resolution analysis is added to a reproducible Report. The README also documents JSON CLI output and standalone structure reading. Source excerpts show the core orchestration and reader paths; relationships not evidenced there are omitted.
Main entry points: modelgate/__init__.py Python API · manifest.py Normalized Manifest · _checkers/__init__.py Normative checker registry · report.py Audit report
Components 17
Entry points
- Python API
packages/modelgate-core/src/modelgate/__init__.py - Command-line interface
packages/modelgate-core/src/modelgate/cli.py
Dataset ingestion
- Reader dispatch
packages/modelgate-core/src/modelgate/_readers/__init__.py - ZIP reader
packages/modelgate-core/src/modelgate/_readers/zip_reader.py - ImageFolder reader
packages/modelgate-core/src/modelgate/_readers/imagefolder_reader.py - Structure detection
packages/modelgate-core/src/modelgate/_readers/_structure.py - Normalized Manifest
packages/modelgate-core/src/modelgate/manifest.py
MGS evaluation
- Normative checker registry
packages/modelgate-core/src/modelgate/_checkers/__init__.py - Structure check
packages/modelgate-core/src/modelgate/_checkers/structure.py - Integrity check
packages/modelgate-core/src/modelgate/_checkers/integrity.py - Duplicate check
packages/modelgate-core/src/modelgate/_checkers/duplicate.py - Balance check
packages/modelgate-core/src/modelgate/_checkers/balance.py - Resolution analysis
packages/modelgate-core/src/modelgate/_checkers/resolution.py - Spec rounding
packages/modelgate-core/src/modelgate/_rounding.py
Results
- Audit report
packages/modelgate-core/src/modelgate/report.py
Other components
- Dataset user
- Dataset
Connections 34
- Dataset user → Python API: calls audit
README.md - Dataset user → Command-line interface: runs check
README.md - Dataset → Python API: supplies path
README.md - Dataset → Command-line interface: supplies path
README.md - Command-line interface → Python API: invokes API
packages/modelgate-core/src/modelgate/cli.py - Python API → Reader dispatch: reads dataset
packages/modelgate-core/src/modelgate/__init__.py - Reader dispatch → ZIP reader: selects reader
packages/modelgate-core/src/modelgate/_readers/__init__.py - Reader dispatch → ImageFolder reader: selects reader
packages/modelgate-core/src/modelgate/_readers/__init__.py - ZIP reader → Structure detection: classifies paths
packages/modelgate-core/src/modelgate/_readers/zip_reader.py - ImageFolder reader → Structure detection: classifies paths
packages/modelgate-core/src/modelgate/_readers/imagefolder_reader.py - ZIP reader → Normalized Manifest: builds manifest
packages/modelgate-core/src/modelgate/_readers/zip_reader.py - ImageFolder reader → Normalized Manifest: builds manifest
packages/modelgate-core/src/modelgate/_readers/imagefolder_reader.py - Python API → Normative checker registry: gets checkers
packages/modelgate-core/src/modelgate/__init__.py - Python API → Normalized Manifest: passes manifest
packages/modelgate-core/src/modelgate/__init__.py - Python API → Resolution analysis: computes resolution
packages/modelgate-core/src/modelgate/__init__.py - Python API → Audit report: constructs report
packages/modelgate-core/src/modelgate/__init__.py - Normative checker registry → Structure check: registers checker
packages/modelgate-core/src/modelgate/_checkers/__init__.py - Normative checker registry → Integrity check: registers checker
packages/modelgate-core/src/modelgate/_checkers/__init__.py - Normative checker registry → Duplicate check: registers checker
packages/modelgate-core/src/modelgate/_checkers/__init__.py - Normative checker registry → Balance check: registers checker
packages/modelgate-core/src/modelgate/_checkers/__init__.py - Structure check → Normalized Manifest: reads manifest
packages/modelgate-core/src/modelgate/_checkers/structure.py - Integrity check → Normalized Manifest: reads manifest
packages/modelgate-core/src/modelgate/_checkers/integrity.py - Duplicate check → Normalized Manifest: reads manifest
packages/modelgate-core/src/modelgate/_checkers/duplicate.py - Balance check → Normalized Manifest: reads manifest
packages/modelgate-core/src/modelgate/_checkers/balance.py - Resolution analysis → Normalized Manifest: reads manifest
packages/modelgate-core/src/modelgate/_checkers/resolution.py - Integrity check → Audit report: returns result
packages/modelgate-core/src/modelgate/_checkers/integrity.py - Duplicate check → Audit report: returns result
packages/modelgate-core/src/modelgate/_checkers/duplicate.py - Balance check → Audit report: returns result
packages/modelgate-core/src/modelgate/_checkers/balance.py - Structure check → Audit report: returns result
packages/modelgate-core/src/modelgate/_checkers/structure.py - Balance check → Spec rounding: rounds metrics
packages/modelgate-core/src/modelgate/_checkers/balance.py - Duplicate check → Spec rounding: rounds metrics
packages/modelgate-core/src/modelgate/_checkers/duplicate.py - Integrity check → Spec rounding: rounds metrics
packages/modelgate-core/src/modelgate/_checkers/integrity.py - Resolution analysis → Spec rounding: rounds metrics
packages/modelgate-core/src/modelgate/_checkers/resolution.py - Python API → Dataset user: returns report
README.md