Model Tutorials

The OpenVINO™ toolkit supports most TensorFlow and PyTorch models. The following table lists deep-learning models commonly used in the Embodied Intelligence solutions, and information on how to run them on Intel® platforms:

Algorithm	Description	Link
YOLOv8	CNN-based object detection	YOLOv8
YOLOv12	CNN-based object detection	YOLOv12
MobileNetV2	CNN-based object detection	MobileNetV2
SAM	Transformer-based segmentation	SAM
SAM2	Extends SAM to video segmentation and object tracking with cross attention to memory	SAM2
FastSAM	Lightweight substitute to SAM	FastSAM
MobileSAM	Lightweight substitute to SAM (Same model architecture as SAM. See OpenVINO SAM tutorials for model export and application)	MobileSAM
U-NET	CNN-based segmentation and diffusion model	U-NET
DETR	Transformer-based object detection	DETR
GroundingDino	Transformer-based object detection	GroundingDino
CLIP	Transformer-based image classification	CLIP
Qwen2.5VL	Multimodal large language model	Qwen2.5VL
Whisper	Automatic speech recognition	Whisper
FunASR	Automatic speech recognition	FunASR Setup in LLM Robotics - sample pipeline

Attention: When following these tutorials for model conversion, ensure that the OpenVINO toolkit version used for model conversion is the same as the runtime version used for inference. Otherwise, unexpected errors may occur, especially if the model is converted using a newer version and the runtime is an older version. See details in the Troubleshooting section.

Please also find information for the models of imitation learning, grasp generation, simultaneous localization and mapping (SLAM) and bird’s-eye view (BEV):