We are excited to announce the release of the sinapsis-retina-face-trt package, a powerful toolset for real-time facial recognition and verification. This package leverages TensorRT optimization and supports DeepFace and RetinaFace models, enabling developers to easily implement high-performance face detection, embedding extraction, and verification workflows.
Key Features
RetinaFacePytorch
Runs face detection using the RetinaFace model implemented in PyTorch.
Ideal for developers looking to integrate face detection into their applications with minimal setup.
RetinaFacePytorchTRT
A TensorRT-optimized version of RetinaFacePytorch for faster inference.
Provides significant performance improvements for real-time applications.
RetinaFacePytorchTRTTorchOnly
A Torch-TensorRT optimized version of RetinaFace, focusing solely on Torch-TRT acceleration.
Simplifies the integration of optimized models into existing PyTorch workflows.
PytorchEmbeddingSearch
Performs similarity search over a gallery of embeddings.
Enables developers to quickly identify and match faces in a database.
PytorchEmbeddingExtractor
A base template for extracting embeddings from face images.
Provides flexibility for custom embedding extraction workflows.
Facenet512EmbeddingExtractorTRT
Uses TensorRT for fast embedding extraction based on the Facenet512 model.
Optimized for high-performance face recognition tasks.
Facenet512EmbeddingExtractorTRTDev
An alternative version of Facenet512EmbeddingExtractorTRT that converts the model at runtime.
Ideal for developers testing or experimenting with different model configurations.
FaceVerificationFromGallery
Performs face verification by comparing predicted face embeddings with those stored in a gallery file.
Enables seamless integration of face verification into real-world applications.
Web Applications for Face Recognition and Verification
The package also includes web applications for face recognition and verification, providing developers with out-of-the-box solutions for:
Real-time face detection and recognition in web-based environments.
Face verification workflows that integrate seamlessly with the package's embedding extraction and search templates.