Intel® Geti™ 2.0.0 Release: Advancing AI Model Development

Intel® Geti™ 2.0.0 is here. A new release of Intel’s Vision AI software for building computer vision models faster and with less. This release includes several additions for data labeling, model training, and inferencing.

The release is part of Intel’s ongoing commitment to deliver high-quality software. All existing customers are invited to upgrade to Intel Geti 2.0.0 and take advantage of the latest functionality!

Intel Geti software is part of Intel® Tiber™ Edge Platform a solution enabling enterprises to develop, deploy, run, manage, and scale edge AI applications with cloud-like simplicity, leveraging an unmatched partner ecosystem.​ Intel Geti software is a crucial component of the Edge Platform utilized for building vision AI models.

New Capabilities Spotlight

The latest version of the software, Intel Geti 2.0.0, brings several new features and enhancements, including these highlights:

Live Data Capture and Inferencing Support

• Pull current data from external cameras into the platform for data labeling, model training, or live inference testing.

Advance Data Annotation and Model Training Capabilities

• Convert data annotations to other computer vision training tasks (i.e., from anomaly classification to single-label classification or segmentation to object detection) for easy dataset adaptation and project experimentation.
• Achieve more effective AI models faster by leveraging Vision Transformer (ViT) and new model architectures that boost accuracy and efficiency in training tasks.

Broader GPU Support for Model Training

• New Intel® GPU Max 1100 support.
• Tap into a wider array of GPU choices that meet the memory specifications through flexible GPU compatibility checks.

With Generative AI and foundation models based on transformers taking the world by storm, we enabled Intel Geti users to take advantage of the vision transformer capabilities, such as those built into the Segment Anything Model, to accelerate data labeling and annotation. In this release, we are adding capabilities for users to further fine-tune models based on vision transformer architectures on their custom data. Several new neural network and vision transformer-based architectures are supported in this release, including Swin Transformers, data-efficient image transformers, YOLOX models in various versions, and SegNeXt model architectures.

Additional updates in Intel Geti 2.0.0 include new adjustable parameters for improved segmentation of small objects, additional data import capabilities for COCO, YOLOv8, and VOC formats, and more. Review the product release notes for a complete overview:

Try Intel Geti Software

Free software trials are available for prospective customers. Trialists can access all the software features and functionality for a complete product evaluation.

As part of the Intel Geti 2.0.0 release, the Python SDK functionality available via the Python Package Index “PyPi” is not only accessible for current users but also for trial users evaluating the software. The SDK is a core competence that helps simplify and automate development pipelines with project creation, data and model download, and deployment capabilities.