'StreamPoseML' An End-to-End Open-Source Web Application and Python Toolkit for Real-Time Video Pose Classification and Machine Learning - v0.2.2 - Build and Deployment Fixes

🔧 Bug Fixes & Improvements

This patch release addresses critical build and deployment issues discovered in v0.2.1.

Fixed Issues

• Docker Build Context: Corrected Docker build contexts to work with both manual builds and docker-compose.local.yml
• CI/CD Workflows: Fixed PyPI publishing workflow by adding missing build and twine dependencies
• Code Quality: Resolved linting errors (whitespace on blank lines) and mypy type checking issues
• Volume Pathing: Fixed various volume and path mapping issues in Docker configurations

Docker Images

All Docker images have been rebuilt and pushed to DockerHub with the latest fixes: - mrilikecoding/stream_pose_ml_api:latest - mrilikecoding/stream_pose_ml_web_ui:latest
- mrilikecoding/stream_pose_ml_mlflow:latest

Deployment

• ✅ All CI/CD workflows now pass successfully
• ✅ Docker images are compatible with both local development and production deployment
• ✅ Build process maintains compatibility with existing docker-compose configurations

Breaking Changes

None - this is a patch release with backward compatibility maintained.

Next Steps

See Issue #30 for upcoming enhancements to model loading progress streaming.

Scientific Software - Peer-reviewed - Python
Published by mrilikecoding 4 months ago

'StreamPoseML' An End-to-End Open-Source Web Application and Python Toolkit for Real-Time Video Pose Classification and Machine Learning - v 0.2.1 Refactors

1. Documentation improvements:
   • Updated MLFlow documentation
   • Improved startup process documentation
   • Switched to Alabaster docs theme
   • General documentation updates
2. MLFlow compatibility:
   • Pinned MLFlow to version < 2.22.1 to avoid breaking changes in version 3 and subsequent changes to model serving (gunicorn to FastAPI). Models logged with MLFlow 2.22.1 and later will use FastAPI by default when being served. But if we use a later version of MLFlow with models logged on a version before 2.22.1 this will not work. A later version of this package will switch to only supporting models logged with version 2.22.1 or higher.
   • Removed old MLFlow Dockerfile and run scripts from package code
3. Code refactoring:
   • Separated API service from core package code
   • Restructured imports for better organization
   • Merged test branch with comprehensive testing improvements
   • Added tests for geometry, services, and jobs modules
4. Build and CI improvements:
   • Added Black code formatter to CI flow
   • Fixed API CI path issues
   • Moved API tests to proper directory
   • Improved start/stop process with new make targets for testing
   • Updated build scripts

Scientific Software - Peer-reviewed - Python
Published by mrilikecoding 6 months ago

'StreamPoseML' An End-to-End Open-Source Web Application and Python Toolkit for Real-Time Video Pose Classification and Machine Learning - v0.2.0 (alpha) MLFlow Integration

This release integrates MLFlow support into the web application as well as other MLFlow related support in the pip pacakge. While StreamPoseML provides some specific wrappers around various ScikitLearn features and XGBoost, the idea with MLFlow integration is to enable maximal flexibility around a standard and commonly used toolkit that can be dropped into the web application provided models are trained using an schema consistent with StreamPoseML's mediapipe keypoint generation.

Scientific Software - Peer-reviewed - Python
Published by mrilikecoding about 1 year ago

'StreamPoseML' An End-to-End Open-Source Web Application and Python Toolkit for Real-Time Video Pose Classification and Machine Learning - v.0.1.1 Build updates

This update adds some build configuration

Scientific Software - Peer-reviewed - Python
Published by mrilikecoding about 2 years ago

'StreamPoseML' An End-to-End Open-Source Web Application and Python Toolkit for Real-Time Video Pose Classification and Machine Learning - v0.1.0 Initial Alpha Release

Initial alpha release. This release represents the initial work of pulling this package together as a usable pip package alongside a parallel web application stack, as well as example notebooks from the initial use case.

Scientific Software - Peer-reviewed - Python
Published by mrilikecoding about 2 years ago