OpenVINO™ Training Extensions is a low-code transfer learning framework for Computer Vision. The CLI commands of the framework allows users to train, infer, optimize and deploy models easily and quickly even with low expertise in the deep learning field. OpenVINO™ Training Extensions offers diverse combinations of model architectures, learning methods, and task types based on PyTorch and OpenVINO™ toolkit.
OpenVINO™ Training Extensions provides a "model template" for every supported task type, which consolidates necessary information to build a model. Model templates are validated on various datasets and serve one-stop shop for obtaining the best models in general. If you are an experienced user, you can configure your own model based on torchvision, pytorchcv, mmcv and OpenVINO Model Zoo (OMZ).
Furthermore, OpenVINO™ Training Extensions provides automatic configuration of task types and hyperparameters. The framework will identify the most suitable model template based on your dataset, and choose the best hyperparameter configuration. The development team is continuously extending functionalities to make training as simple as possible so that single CLI command can obtain accurate, efficient and robust models ready to be integrated into your project.
OpenVINO™ Training Extensions supports the following computer vision tasks:
- Classification, including multi-class, multi-label and hierarchical image classification tasks.
- Object detection including rotated bounding box support
- Semantic segmentation
- Instance segmentation including tiling algorithm support
- Action recognition including action classification and detection
- Anomaly recognition tasks including anomaly classification, detection and segmentation
OpenVINO™ Training Extensions supports the following learning methods:
- Supervised, incremental training, which includes class incremental scenario and contrastive learning for classification and semantic segmentation tasks
- Semi-supervised learning
- Self-supervised learning
OpenVINO™ Training Extensions will provide the following features in coming releases:
- Distributed training to accelerate the training process when you have multiple GPUs
- Half-precision training to save GPUs memory and use larger batch sizes
- Integrated, efficient hyper-parameter optimization module (HPO). Through dataset proxy and built-in hyper-parameter optimizer, you can get much faster hyper-parameter optimization compared to other off-the-shelf tools. The hyperparameter optimization is dynamically scheduled based on your resource budget.
- OpenVINO™ Training Extensions uses Datumaro as the backend to hadle datasets. Thanks to that, OpenVINO™ Training Extensions supports the most common academic field dataset formats for each task. We constantly working to extend supported formats to give more freedom of datasets format choice.
- Auto-configuration functionality. OpenVINO™ Training Extensions analyzes provided dataset and selects the proper task and model template to provide the best accuracy/speed trade-off. It will also make a random auto-split of your dataset if there is no validation set provided.
Please refer to the installation guide for installing it using Python venv (virtual environment), or use the following steps to run it in a containerised environment:
$ docker build \
-t trainer \ # image tag, required
--build-arg UBUNTU_VER=20.04 \ # default Ubunutu version, optional
--build-arg PYTHON_VER=3.9 \ # default Python version, optional
--build-arg SOURCE=https://download.pytorch.org/whl/cpu \ # default (CPU) deps, optional
. # training_extensions/
$ docker run \
-it \ # enter interactive terminal
--rm \ # remove container after use
-v "$(pwd)/shared:/mnt/shared:rw" \ # shared volume to host machine
--shm-size=4g \ # increase mounted shared memory
trainer
trainer$ otx # ... installed on Ubuntu 20.04 with /mnt/shared as shared directory
Note: Python 3.8 and 3.9 were tested, along with Ubuntu 18.04 and 20.04.
otx find
helps you quickly find the best pre-configured models templates as well as a list of supported backbonesotx build
creates the workspace folder with all necessary components to start training. It can help you configure your own model with any supported backbone and even prepare a custom split for your datasetotx train
actually starts training on your datasetotx eval
runs evaluation of your trained model in PyTorch or OpenVINO™ IR formatotx optimize
runs an optimization algorithm to quantize and prune your deep learning model with help of NNCF and POT tools.otx export
starts exporting your model to the OpenVINO™ IR formatotx deploy
outputs the exported model together with the self-contained python package, a demo application to port and infer it outside of this repository.otx demo
allows one to apply a trained model on the custom data or the online footage from a web camera and see how it will work in a real-life scenario.otx explain
runs explain algorithm on the provided data and outputs images with the saliency maps to show how your model makes predictions.
You can find more details with examples in the CLI command intro.
- Support direct annotation input for COCO format (openvinotoolkit#1921)
- Action task supports multi GPU training. (openvinotoolkit#2057)
- Support storage cache in Apache Arrow using Datumaro for action tasks (openvinotoolkit#2087)
- Add a simplified greedy labels postprocessing for hierarchical classification (openvinotoolkit#2064).
- Support auto adapting batch size (openvinotoolkit#2119)
- Support auto adapting num_workers (openvinotoolkit#2165)
Please refer to the CHANGELOG.md
- develop
- Mainly maintained branch for developing new features for the future release
- misc
- Previously developed models can be found on this branch
OpenVINO™ Toolkit is licensed under Apache License Version 2.0. By contributing to the project, you agree to the license and copyright terms therein and release your contribution under these terms.
Please use Issues tab for your bug reporting, feature requesting, or any questions.
misc branch contains training, evaluation, and export scripts for models based on TensorFlow and PyTorch. These scripts are not ready for production. They are exploratory and have not been validated.