sinapsis-object-detection v0.4.2

Updated dfine to use models from HF

sinapsis-object-detection v0.1.0

In this release we present an update to sinapsis-object-detection by including the sinapsis-ultralytics package with support for the Ultralytics library.

The sinapsis-ultralytics package provides a comprehensive suite of tools for working with Ultralytics models, that can be easily integrated in the sinapsis framework. This release introduces four specialized templates—UltralyticsTrain, UltralyticsVal, UltralyticsPredict, and UltralyticsExport—designed to optimize workflows for training, validating, deploying, and exporting models. These templates empower developers to efficiently integrate and deploy Ultralytics models into their applications.

Key Features

UltralyticsTrain

Template for training Ultralytics models.
Supports various model configurations and training parameters.
Simplifies the process of fine-tuning models for specific use cases.

UltralyticsVal

Template for validating Ultralytics models.
Enables accurate evaluation of model performance on validation datasets.
Provides detailed metrics and insights for model optimization.

UltralyticsPredict

Template for generating inference predictions with trained models.
Supports real-time and batch prediction workflows.
Ideal for deploying models in production environments.

UltralyticsExport

Template for exporting models to deployment-ready formats.
Facilitates the conversion of trained models for use in various deployment scenarios.
Simplifies the process of preparing models for inference in production.

The package also includes webapps that showcase the functionality of UltralyticsTrain and UltrlayticsInference for classification, detection, segmentation, Pose detection and Oriented Object detection tasks

sinapsis-object-detection v0.2.0

sinapsis-object-detection mono-repo continues to provide powerful tools and utilities to streamline the configuration and deployment of object detection pipelines—helping developers build advanced AI-driven applications with ease.

What's New

• Expanded Template Support
  The mono-repo now includes the sinapsis-rfdetr package with support for the RFDETR object detection library, and utilities to simplify the configuration and deployment of advanced object detection systems, helping developers go from prototype to production with ease.

Key Features:

• RFDETRExport & RFDETRLargeExport

      Templates for exporting RFDETRBase and RFDETRLarge models to ONNX—streamlining model deployment across platforms.
  

• RFDETRInference & RFDETRLargeInference

      Inference templates for evaluating RFDETR models on image datasets.
  

• RFDETRTrain & RFDETRLargeTrain

      Training templates for both RFDETR variants, optimized for quick experimentation and scalability.
  

What’s Next?

The roadmap for sinapsis-object-detection includes:

• Additional object detection templates to support more frameworks and models.
• Native support for multi-label object detection, unlocking more complex detection scenarios.

With these updates, sinapsis-object-detection continues to evolve into a robust, modular, and developer-friendly framework for modern object detection workflows.

sinapsis-object-detection v0.1.0

We are excited to announce the release of sinapsis-object-detection, a powerful mono-repository designed to simplify the development and deployment of advanced object detection systems, and integrate it with the sinapsis platform.

• The sinapsis-object-detection mono-repo provides a suite of templates and utilities to streamline the configuration and operation of object detection models. It is designed to help developers quickly implement and experiment with cutting-edge object detection pipelines and integrate with the sinapsis platform
• sinapsis-dfine package: a specialized module designed to simplify the deployment of object detection systems using the D-FINE model. Key features include:

DFINETraining Template:
A ready-to-use template for the D-FINE model's training pipeline. It streamlines the setup process by handling configuration initialization, weight downloading, and training solver configuration. This enables developers to quickly start training advanced object detection models with minimal setup.
DFINEInference Template:
A ready-to-use template for performing inference on a set of images using the different D-FINE architectures available. This template enables developers to quickly experiment with various D-FINE configurations and deploy them in real-world applications.

• The sinapsis-object-detection mono repo is a game-changer for developers working on advanced object detection tasks. By providing a unified, modular framework for building and deploying object detection systems, it reduces the complexity of getting started with AI-driven applications.

• The addition of the sinapsis-dfine package further strengthens the package’s versatility, offering developers a powerful toolset for experimentation and deployment.

What’s Next?
We’re already planning future updates to expand the capabilities of the sinapsis-object-detection package. Upcoming features will include:

Additional templates for advanced object detection functionalities.
Support for multi-label detection: Stay tuned for enhancements that will enable developers to handle more complex object detection tasks.
Integration with popular object detection datasets and benchmarks: Simplifying the process of evaluating and comparing different object detection models.