OpenVINO™ Training Extensions is a low-code transfer learning framework for Computer Vision. The API & 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 "recipe" 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.
Starting with OTX v2.4.5, we introduced a new repository structure and a more flexible backend concept. We're excited to present support for multiple backends — beginning with the OpenVINO™ backend, while all previous OTX functionality is now organized under the "native" backend.
In the future, we plan to integrate popular third-party libraries such as Anomalib <https://github.com/open-edge-platform/anomalib>_, Transformers <https://huggingface.co/docs/transformers/index>_, and more — seamlessly integrated into the repository.
This will enable users to train, test, export, and optimize a wide variety of models from different backends using the same CLI commands and unified API, without the need for reimplementation.
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 and tiling support
- Semantic segmentation including tiling algorithm support
- Instance segmentation including tiling algorithm support
- Anomaly recognition tasks including anomaly classification, detection and segmentation
OpenVINO™ Training Extensions provides the following usability features:
- Native Intel GPUs (XPU) support. OpenVINO™ Training Extensions can be installed with XPU support to utilize Intel GPUs for training and testing.
- Datumaro data frontend: OpenVINO™ Training Extensions supports the most common academic field dataset formats for each task. We are constantly working to extend supported formats to give more freedom of datasets format choice.
- Distributed training to accelerate the training process when you have multiple GPUs
- Mixed-precision training to save GPUs memory and use larger batch sizes
- Class incremental learning to add new classes to the existing model
- Model deployment to OpenVINO™ IR and ONNX formats and inference with OpenVINO™ ModelAPI
- Multiple backend support to easily adapt models from third-party implementations into the OpenVINO™ Training Extensions repository.
Please refer to the installation guide. If you want to make changes to the library, then a local installation is recommended.
Install from PyPI
Installing the library with uv tool is the easiest way to get started with otx.uv pip install otx[cuda]For Intel GPUs users:
uv pip install otx[xpu]Install from source
To install from source, you need to clone the repository and install the library using uv pip via editable mode.# Create a new virtual environment using uv (Python 3.11)
uv venv .otx --python 3.11
source .otx/bin/activate
# Clone the repository
git clone https://github.com/open-edge-platform/training_extensions.git
cd training_extensions
# Install in editable mode
uv pip install -e .[cuda]
# For Intel GPUs users
uv pip install -e .[xpu]OpenVINO™ Training Extensions supports both API and CLI-based training. The API is more flexible and allows for more customization, while the CLI training utilizes command line interfaces, and might be easier for those who would like to use OpenVINO™ Training Extensions off-the-shelf.
For the CLI, the commands below provide subcommands, how to use each subcommand, and more:
# See available subcommands
otx --help
# Print help messages from the train subcommand
otx train --help
# Print help messages for more details
otx train --help -v # Print required parameters
otx train --help -vv # Print all configurable parametersYou can find details with examples in the CLI Guide. and API Quick-Guide.
Below is how to train with auto-configuration, which is provided to users with datasets and tasks:
API Usage
from otx.engine import create_engine
# get all the available recipes for all tasks
from otx.backend.native.cli.utils import list_models
model_lists = list_models(print_table=True)
# instantiate native otx engine with atss model for object detection
engine = create_engine(data="path/to/dataset/root", model="src/otx/recipe/detection/atss_mobilenetv2.yaml")
engine.train()
engine.test()
exported_path = engine.export()
# by default all artifacts are stored in "./otx-workspace" directory.
# working directory can be specified
engine = create_engine(data="path/to/dataset/root", model="src/otx/recipe/detection/atss_mobilenetv2.yaml", work_dir="my_workdir")
# openvino backend is used to validate and optimize exported OpenVINO IR models
ov_engine = create_engine(data="path/to/dataset/root", model=exported_path)
ov_engine.test()
ov_engine.optimize()For more examples, see documentation: API Quick-Guide
CLI Usage
# get all recipes list
otx find
# otx train
otx train --config src/otx/recipe/detection/atss_mobilenetv2.yaml --data_root data/wgisd
# by default, working directory is "./otx-workspace". It can be specified with "--work_dir" parameter
otx test --config src/otx/recipe/detection/atss_mobilenetv2.yaml --data_root data/wgisd --checkpoint otx-workspace/.latest/train/best_checkpoint.ckpt
otx export --config src/otx/recipe/detection/atss_mobilenetv2.yaml --data_root data/wgisd --checkpoint otx-workspace/.latest/train/best_checkpoint.ckpt
# or using work_dir
otx test --work_dir otx-workspace/.latest/train
otx export --work_dir otx-workspace/.latest/train
# directly from working directory
cd otx-workspace
otx test
otx export
For more examples, see documentation: CLI Guide
In addition to the examples above, please refer to the documentation for tutorials on using custom models, training parameter overrides, and tutorial per task types, etc.
Please refer to the CHANGELOG.md
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.
Intel is committed to respecting human rights and avoiding complicity in human rights abuses. See Intel's Global Human Rights Principles. Intel's products and software are intended only to be used in applications that do not cause or contribute to a violation of an internationally recognized human right.
For those who would like to contribute to the library, see CONTRIBUTING.md for details.
Thank you! we appreciate your support!