Would like to share that we developed a PI Ai Hat in the form factor of a Raspberry PI Zero. More information can be found on the following link.
The XAPIZ3500 is based on our AI module XAM3500. To showcase our Machine learning accelerator module, we created a Raspberry PI Hat the size of a PI Zero. Coupled that with 3D Printing Technology, you can now build your own AI Camera.
It allows you to add Machine Learning accelerator to your Raspberry PI empowering you to create many interesting applications. Adding the XAPIZ3500 allows Yolo-Like object detection with close to 10X performance improvements.
The bandwidth between PI and XAM3500 is 40Mbps. The theoretical bandwidth is about 20FPS for QVGA color video frame.
YoloV2 Performance (current state…. can be further optimized )
Pi Zero ≈ 3FPS
Pi 3B+ ≈ 8FPS
We like the form factor of Raspberry PI Zero but we find it a little underpowered when running our TensorFlow model. That is why we decided to design an AI Hat that offloads the inference engine running our TensorFlow models.
While there are other solutions such as adding a USB based neural engine accelerator, it spoils the beauty of the PI Zero form factor.