Intel® Geti™ 1.2.0#
Release Summary#
Intel® Geti™ 1.2.0 contains several updates and feature enhancements, including key highlights:
Customizable tiling algorithm - Boost accuracy by dividing high-resolution images into smaller tiles during pre-processing for small object detection and counting tasks.
View training, test, and validation subsets - Recognize what datasets a model is training from and identify opportunities to optimize results.
Support for OpenVINO 2022.3 Long Term Support (LTS) - Tap into the new OpenVINO 2022.3 functionality and performance.
Model architecture switching - Switch the active model to a different architecture without completely retrain it.
Support for Datumaro format - Export and import your datasets in the Datumaro format for flexible label relations.
Responsive UI design - Responsive design is now available for more devices and screen sizes.
Support for the latest Ubuntu version 22.04 LTS - Now you can install the Intel® Geti™ platform on the latest version of Ubuntu to date.
Release Details#
This section covers additional details on the new functionality available with Intel® Geti™ 1.2.0.
Customizable tiling algorithm#
A customizable tiling algorithm is now available for small object detection and counting. Boost accuracy by dividing high-resolution images into smaller tiles during pre-processing.
View training, test, and validation subsets#
You can examine the training, test, and validation subsets as thumbnails in the Models section of each project and for each model version. This feature shows the exact datasets used for model training and can help you to identify opportunities to improve your model.
To better understand what data was used in each phase of the training process, read more about the dataset split in this section.
Support for OpenVINO 2022.3 LTS#
The Intel® Geti™ platform supports OpenVINO 2022.3 LTS, the latest version of OpenVINO to date. You can export optimized models for deployment with OpenVINO 2022.3 to run on the latest Intel hardware.
Model architecture switching#
This release introduces active model switching between multiple architectures*. You can easily switch between different model architectures while keeping the model parameters configuration intact (from the previous active model). This feature allows for more efficient model training as there is no need to retrain the model when switching the architecture.
Disclaimer: Model switching between versions of the same architecture is not yet supported.
Support for Datumaro dataset format#
Now, you can export and import datasets in a Datumaro format, which stores annotations in a JSON file, allowing for highly flexible label relations. This feature is available for segmentation, detection, and single-label classification tasks. However, support for a hierarchical label structure and video data export are not available.
Responsive UI design#
The Intel® Geti™ platform now provides a responsive design across devices and screen sizes supporting 1024px or higher resolutions. This update ensures all content is accessible and optimized for the user’s device of choice, while considering the varying screen resolutions, orientations, and reducing cognitive load while navigating the platform.
Dataset view at 768px width |
Annotation view at 768px width |
|---|---|
|
|
Support for the latest Ubuntu version 22.04 LTS#
The Intel® Geti™ platform now supports the latest major version of Ubuntu, version 22.04.1 LTS.
System Requirements for Intel® Geti™ 1.2.0#
The platform can be installed on a machine with the following minimum hardware requirements:
CPU for workstations: Intel® Core™ i7, Intel® Core™ i9 or Intel® Xeon® scalable processors family capable of running 20 concurrent threads (in case of using the default K3s) or 48 concurrent threads (in case of using pre-installed K8s).
Note
From Intel® Core™ family, we recommend the following CPU series:
13th gen (Raptor Lake): Intel® Core™ i7 13700 series and Intel® Core™ i9 13900 series
12th gen (Alder Lake): Intel® Core™ i9 12900 series
CPU for cloud deployments: CPUs capable of running min. 24 concurrent threads for K3s or min. 48 concurrent threads for K8s (so for example, on AWS EC2 instances, this requirement would be translated to min. 24 vCPUs for K3s or min. 48 vCPUs for K8s).
GPU: min. one NVIDIA GPU with min. 16GB of memory (e.g. RTX 3090, RTX 6000, RTX 8000, Tesla A100, Tesla V100, Tesla P100, or Tesla T4), other NVIDIA GPUs in a similar series are likely also compatible if they meet minimum memory requirements. However, the full range of devices is not fully tested and not specifically supported. We recommend 24GB of memory for the stable training & optimization.
Memory: min. 64 GB RAM (128 GB recommended)
Disk Space: min. 500 GB (1 TB recommended) for a root partition
Installation of the platform on a multinode configuration of Kubernetes cluster is not supported.