Following the successful launch of the Raspberry Pi AI Kit and AI Camera, we are thrilled to unveil the newest addition to our AI product line: the Raspberry Pi AI HAT+. Designed to enhance your AI projects, the AI HAT+ builds on our commitment to delivering cutting-edge technology at an accessible price.
The AI HAT+ features best-in-class Hailo AI accelerator technology, just like our AI Kit, but with an exciting new twist: two performance options. Choose between the 13 TOPS (tera-operations per second) model, equipped with the same Hailo-8L accelerator as the AI Kit, or opt for the more powerful 26 TOPS model, which comes with the advanced Hailo-8 accelerator.
Both versions are designed to comply with our HAT+ specification, enabling the AI HAT+ to automatically switch to PCIe Gen 3.0 mode, maximizing the 26 TOPS of compute power. This ensures that you can fully harness the capabilities of the Hailo-8 accelerator.
One of the standout features of the AI HAT+ is its integration of the Hailo accelerator chip directly onto the main PCB. Unlike the AI Kit, which uses an M.2 connector, this integration simplifies setup and improves thermal dissipation. As a result, the AI HAT+ can handle demanding AI workloads more efficiently, making it an ideal choice for projects that require robust performance.
So, what can you achieve with the 26 TOPS model over the 13 TOPS model? Simply put, more! The increased processing power allows you to run more sophisticated neural networks in real-time, resulting in better inference performance. You can execute multiple networks simultaneously at high frame rates. For example, imagine performing object detection, pose estimation, and subject segmentation simultaneously on a live camera feed—now made possible with the 26 TOPS AI HAT+.
Both versions of the AI HAT+ maintain full backward compatibility with the AI Kit. Our existing integration of the Hailo accelerator in the camera software stack operates seamlessly with the AI HAT+. This means any neural network model compiled for the Hailo-8L will run smoothly on the Hailo-8. Although models specifically built for the Hailo-8 may not work on the Hailo-8L, alternative versions with lower performance are generally available, ensuring flexibility for various use cases.
After an exciting few months of AI product releases, we now offer an extensive range of options for running inferencing workloads on Raspberry Pi. Many workloads—particularly those that are sparse, quantized, or intermittent—run natively on Raspberry Pi platforms. For more demanding tasks, we strive to be the best possible embedded host for accelerator hardware like our AI Camera and the new AI HAT+.
We can’t wait to see what you’ll create with the Raspberry Pi AI HAT+. Whether you're a hobbyist, educator, or professional developer, the possibilities are endless. Dive into the world of AI with us and elevate your projects to new heights!
Leave a comment