Platform and Hardware Overview ¶

This page provides an overview of AI model support across different hardware accelerators for each version of CVEDIA-RT.

Models ¶

The following AI models are covered in the tables below:

Detection of people, vehicles, and animals. Used in standard analytics and tracking. Does not include attribute extraction.

Enables Standard Security features such as: intrusion, loitering, crowding, tailgating, object guarding, object left behind, and line crossing.

Extraction of visual attributes such as age, gender, colors, and other appearance details.

Also powers Enhanced Search and enables premium analytics including Armed Person and Fallen Person detection.

Recognition and parsing of vehicle license plates by reading alphanumerical symbols using OCR technology.

Real-time crowd density estimation.

Not supported in the CVEDIA-RT plugin for VMS platforms (e.g., Milestone, Nx Witness).

Extraction of person and object features to enable cross-camera re-identification.

Not included in the CVEDIA-RT plugin for VMS platforms (e.g., Milestone, Nx Witness).

The latest version of CVEDIA-RT supports the following AI models across hardware platforms:

AI Model	NVIDIA	Intel	Hailo	DeepX	Blaize	Ambarella	Rockchip
Detection	✓	✓	✓	✓	✓	✓	✓
Attribute Extraction	✓	✓	✓	✓	✓	✗	✓
LPR	✓	✓	✗	✗	✗	✗	✗
Crowd Estimation	✓	✓	✓	✗	✓	✗	✓
ReID	✓	✓	✓	✗	✓	✗	✗

The 2024.x release series supports the following AI models:

AI Model	NVIDIA	Intel	Hailo	DeepX	Blaize	Ambarella	Rockchip
Detection	✓	✓	✓	✓	✓	✓	✓
Attribute Extraction	✓	✓	✗	✓	✗	✗	✗
Crowd Estimation	✓	✓	✗	✗	✗	✗	✗

Earlier versions have more limited model support:

AI Model	NVIDIA	Intel	Hailo	DeepX	Blaize	Ambarella	Rockchip
Detection	✓	✓	✓	✓	✓	✓	✓
Crowd Estimation	✓	✓	✗	✗	✗	✗	✗

For detailed installation and configuration instructions for each platform, see the following guides: