Product BriefProduct Overview
The Raspberry Pi AI HAT+ (26 TOPS) is a high-performance accelerator designed for users who require significant compute power for computer vision tasks. Built around the tried-and-true Hailo-8 processor, this board doubles the inferencing capability of the entry-level model, delivering 26 TOPS of neural processing power to the Raspberry Pi 5.
Why Choose 26 TOPS?
While the 13 TOPS version is well-suited to single tasks, the 26 TOPS variant is designed for complex tasks. It excels in scenarios where you need to run multiple neural networks simultaneously or process high-resolution video streams at higher frame rates.
- Multi-Model Pipelines: Run an object detector, a pose estimator, and a face recognizer simultaneously without dropping frames.
- Low Latency: While it uses less power than its larger siblings, 13 TOPS is more than sufficient to run complex neural networks for object detection, semantic segmentation, and pose estimation in real time.
Seamless PCIe Integration
Gone are the days of USB accelerators hanging off the side of your board. The AI HAT+ communicates directly with the Raspberry Pi 5 over the PCIe Gen 3 interface.
- Auto-Detection: On the latest Raspberry Pi OS, the board is automatically detected, and the NPU is immediately available for tasks.
- Native Camera Support: The built-in
rpicam-appssuite natively supports this module, enabling your camera pipelines to automatically offload post-processing tasks to the NPU with minimal configuration.
Optimized for Complexity
The true strength of the 26 TOPS variant lies in its ability to handle multi-model pipelines. Unlike the 13 TOPS version, this accelerator has the muscle to run multiple neural networks simultaneously without dropping frames. You can run an object detector, a pose estimator, and a face recognizer all at the same time on a single video feed. This capability is essential for advanced projects that require a comprehensive understanding of the environment, not just simple object identification.
