Key Features

What you can do

Computer Vision Task Support

Supports classification (multi-class, multi-label, hierarchical), object detection (including rotated bounding boxes and tiling), semantic and instance segmentation, and anomaly recognition tasks.

Native Intel GPU (XPU) Support

Enables training and testing directly on Intel GPUs without additional setup.

Distributed and Mixed-Precision Training

Supports distributed training across multiple GPUs and mixed-precision training to handle larger batch sizes efficiently.

Validated Model Templates

Provides pre-configured model templates tested on various datasets to facilitate reliable model training.

Model Export and Backend Support

Exports trained models to OpenVINO IR and ONNX formats and supports multiple backends for inference and third-party model adaptation.

Incremental Learning

Allows adding new classes to pre-trained models without retraining from scratch.