Training Toolbox for PyTorch* provides a convenient environment to train Deep Learning models and convert them using OpenVINO™ Toolkit for optimized inference.
- Ubuntu* 16.04 / 18.04
- Python* 3.6+
- libturbojpeg
- For Python prerequisites, refer to
requirements.txt
- (Optional) OpenVINO™ R3 to export trained models
After installation, you are ready to train your own models, evaluate and use them for prediction.
- Action Recognition
- Action Recognition 2
- Image classification
- Object Re-Identification
- Human Pose Estimation
- Instance Segmentation
- Object Detection
- Eye State Classification
- Segmentation of Thoracic Organs
- Super Resolution
- Formula recognition
- Text Spotting
In order to run tests please execute following commands:
python3 tests/run_model_templates_tests.py