This is an Object Detection and Instance Segmentation toolbox, that is a part of OpenVINO™ Training Extensions.
Project is based on open-mmlab's MMDetection.
Please refer to install.md for installation and dataset preparation.
Please see getting_started.md for the basic usage. There are also tutorials for finetuning models, adding new dataset, designing data pipeline, and adding new modules.
This project allows making compression of the models by NNCF framework.
At the moment the following compression methods are supported:
- int8 quantization
- sparsity
- filter pruning
See details in nncf_compression.md.
Please see model_zoo.md for the list of pre-trained models.
We appreciate all contributions to improve this project. Please refer to CONTRIBUTING.md for the contributing guideline.
This repo is currently maintained by Ilya Krylov (@Ilya-Krylov), Pavel Druzhkov (@druzhkov-paul) and Alexander Dokuchaev (@AlexanderDokuchaev), Dmitry Sidnev (@DmitriySidnev).