Geti

Geti is Intel's computer vision platform for building production models without writing training code. It is part of Intel's Open Edge Platform and is aimed at domain experts as much as data scientists: the stated goal is to get from a set of images to a deployable model in minutes and with far less labeled data than a conventional training pipeline demands. The platform covers the full path from data to deployment. Images, video and live camera streams can be ingested, annotation is accelerated by assistive models, and training runs against a catalog of more than 1,600 model variants drawn from 67 architectures. Object detection is the broadest area, with YOLOX, YOLO11, YOLO12 and YOLO26 families alongside real time detection transformers such as D FINE, RF DETR, RT DETR and a DINOv3 based detector. Instance segmentation offers Mask R CNN backbones and segmentation variants of the YOLO families, and image classification spans compact mobile networks, EfficientNet, vision transformers and any timm backbone for fine tuning. Every trained model is exported to OpenVINO intermediate representation in INT8, FP16 or FP32, so the accuracy and latency tradeoff can be chosen for the target device. There are no paid plans and no published prices. Geti is released as open source under the Apache 2.0 license and is installed and run by the user, with three documented routes: prebuilt Docker images for Intel accelerators and NVIDIA CUDA, a native Windows application supplied as an MSIX installer, and running the server and user interface from source for development work. The variant count includes the timm backbones available for classification fine tuning.