Just won a rpi starter kit and now what...

Hi There, I am the lucky winner of a rpi starter kit (13 tops) Unfortunately, I thought there would be a rpi5 included, because now I can’t do anything with it :frowning:
Is it worth buying a rpi5 for this? Is it just to play with or can you actually use it for something in everyday life?

Welcome to the Hailo Community!

Congratulations!

Yes.

A Raspberry Pi is one of the best educational platforms when it comes to computing. With the Raspberry AI HAT you can learn some of the newest technologies in computer vision.
Whether you just want to play and have fun, learn a new skill or develop a new product, is up to you.

What do you do in every day life?

Thank you for answering my questions!

I’m a software developer at a robotics machine builder in the packaging industry, where I’m responsible for the software that handles robot commands and for developing our custom vision application (using MVTec Halcon). Currently, we’re implementing AI within our application, and we’re working on training models based on our clients’ production processes.

This is why I thought this product might be interesting for me. My concern, however, is that it might end up as more of a side project, only to be shelved after a few weeks.

It’s great to have won the hailo-hat, but it would also require purchasing an RPi5, which is another €100,-

I have an rpi3b at home, but it’s mostly gathering dust.

I am not yet completely convinced that I can actually do something with the product besides a video doorbell with recognition. The examples look super nice, but can you also train these types of models yourself?
It is probably not powerful enough to run a personal LLM, for example.

Hi! I am also a software developer at a company that builds robotic machines :slightly_smiling_face:

Edge TPUs like Hailo absolutely have a place in industrial applications. My companies machines have been running AI on TPUs for several years now, I am currently working on porting our models to the Hailo platform to gain some additional throughput from the newer technology.

I have often found though, that having some neat hardware and trying to find a problem is a difficult route and sometimes its hard to come up with a good project. I try to start with the problem to solve and then select the appropriate hardware.

If you want to evaluate Hailo devices for your company then perhaps you can find a small example from your work which could be solved using AI and and a TPU as an alternative to an existing solution? To me LLM’s are unlkely to be a good fit for an EdgeTPU (for now anyway). In my applications I use very small models for very specific tasks which run at very high framerates. A typical model in my system takes between a 256 to 512 pixel wide frame and runs at 5000fps or more on the Hailo8.

2 Likes

For some popular models we provide retraining dockers that make it very easy to retrain models with your own datasets and convert them using the Hailo Model Zoo.

GitHub - Hailo Model Zoo - Retrain on Custom Dataset

For large language models we have designed our next generation accelerator called Hailo-10H.

Hailo-10H M.2 Generative AI Acceleration Module

As you can see from the image the module has a external DDR memory on board. The memory will allow the Hailo-10H to run LLMs which are much larger than CNNs locally on the module without stressing the PCIe interface to the host.

The Hailo-10H is currently not yet generally available. We are working on the software support to ensure we provide a complete solution.

Understand, if you can find a smart kid or student in your family, friends or co-workers, maybe you can give the HAT to them and they can have some fun with it.