Getting access to Himax SDK docs for embedded dev on the Grove Vision AI V2

Hello,

I’m using the Grove Vision AI V2 module and trying to make sense of the board support package and toolchain. I’m a developer and am used to reading and writing C/C++/Makefiles, etc…

I’ve setup the Seeed_Grove_Vision_AI_Module_V2 toolchain from Himax’s github, ran some examples and managed to get the USER LED to blink.

My question is, is there a proper documentation from Himax about their SDK? It seems the only way to learn how to operate the hardware is to try figuring out the sample code and from reading the header files but its far from enough to do any kind of real world embedded project.

For example, trying to figure out sensor_dp_lib just by reading header examples doesn’t make sense and it seems like a crucial component for any kind of serious project involving a camera.

I’ve tried to reach out to Himax to get access to the documentation but sadly no response. Since Seeed partnered with Himax to develop this module I guess you can help in providing detailed documentation.

Appreciate your help!

1 Like

Hi there,
I feel your pain and went down this similar road a while back when it first hit the scene, Sadly IMO
Seeed made a bad choice going with this MCU. I hope that changes but I’m not holding my breath.
Someone was sold a Demo and that’s all they got is what it looks like to me and Others.
my.02
GL :slight_smile: PJ

please keep us in the loop and thanks for contributing your findings. :v:

See if this helps: A collection of useful links for those developing applications for Grove Vision AI V2

I get the impression that Groove Vision AI V2 is a stepping stone to the future Watcher and that Seeed’s focus will be on Watcher afterwards. I tried to get an SDK for the board from HiMax, and they only recommended Seeed’s GitHub. I even gathered some references with the SDK for Arm Cortex M55 and Ethos U5, but I haven’t ventured into studying them yet.

It seems to me that they are trying to deliver a device that is easier to use, supposedly open.

I was awarded a board and have been working with it. There is a step-by-step guide on the wiki that indicates all the steps for a supposed integration with HomeAssistant that never worked.

They are uploading several trained Yolo models, have partnered with NVidia, but there doesn’t seem to be much attention to technical support.

Thx for sharing, I already starred your repo before posting lol.
Unfortunately its not enough for any kind of serious project relying on the HX6538 as the application mcu.

I do believe however that with a rather small effort on Seeed’s part we can get the proper documentation, which must surely exist as some of the header files date back to 2019.
I also don’t see why it is a big issue for Himax either as they have staff that actively maintain the Seeed WE 2 repo, surely they want devs to use their hardware? And its not like we’re asking to open-source the BSP (which is by the way a non issue for companies like Texas Instruments for example), just asking to get the full hardware and BSP manuals, so that we don’t have to guess around and figure out by example that GPIO 20 is the USER LED..

@Seeed ?

If the Himax SDK proves too difficult to navigate, consider looking for alternative SDKs or libraries compatible with the Grove Vision AI hardware.

Hi there,

LOL, I had one of the early releases of the V2 :face_with_hand_over_mouth: SEEED only had a Wiki period to point to, It’s all on SParkFun shoulders outside of that, like toys it is LAME and the stuff is expensive, If you need help they tell you go to sparkfun forum, where it’s banker’s hours for support and only if your thread question is approved (3days)< lame again. :stuck_out_tongue_closed_eyes:They point you to the support page which leads back around in a circle, NOT for beginners either, You better know some stuff before even asking any questions there. Seeed picked them because they have some LARGE claims but little practical delivery. Dev hardware relies on weak ESP12, wifi WTH? the stuff goes belly up if you look at it wrong or make any assumptions about how the Models interact with their hardware. KAKA IMO. documentation is a pile of git hubs. great if you want a migraine :face_with_head_bandage: Himax is NOT my idea of good dev environment, pieces of other stuff all rolled into a YARN ball, navigation is a joke. compared with Nordic whom it’s obvious from the support alone, you can actually call someone and ask a question (a Local Rep in Sarasota) and with out your credit card handy. They want to be in AI more or less only forget the rest of the connectivity world which is vital for success.
Probably why seeed is going in another direction yet again with watcher and abstracting or attempting too add another layer of MUD. Other sdk’s suffer the same fate when it comes to the total integration of BLE, WIFI & cellular IoT they do have a NB-IoT but Azure certification is required to make a products real. No thanks. the Synopsys DesignWare stuff is just a mess all over the place. notice how they ride Google’s coat tails for the good stuff , Google TensorFlow Lite for Microcontrollers (TFLu) “beginners stick with that”
TensorFlow Lite for Microcontrollers very compatible. :+1:
That’s the ONLY promising offering they have links too , IMO.
YMMV
GL :slight_smile: PJ
:v:
The watcher I won in the live stream arrives tomorrow(DHL) looks like a great little package of tech , let’s see how it adds up? Stay tuned.

1 Like

Although I’m still trying to resolve a basic problem with Grove’s compatibility with Home Assistant, I’m still investigating the possibility of creating firmware from scratch for the board. The most plausible clue is the SSCMA-Micro repository, which appears to be the source code of the firmware used in the Grove Vision AI V2 and which possibly contains pinout indications and other information in its code. I had already left it in my notes but I hadn’t checked the source code yet.

But I have found that there is still a lot of work to be done. When I received the board, I started thinking if I could use the SPI bus to use a display, for example. Shortly after, I see the launch of the Watcher, with the same Grove hardware and a display. Let’s see how much it will be possible to customize the software that runs on Seeed’s hardware. For now, there is still a lot of documentation missing.

I had the same idea, as it is basically the only “full” app available for this device.
I’m still expecting Seeed official to comment on this thread and at least provide some input as to what are the plans for this board.

1 Like

I think the folks at Seeed are focusing on the Watcher software, which is the same platform. My impression is that they should publish more advanced API, SDK and tutorials as soon as Watcher is consolidated. But that’s just a guess.

I’ve wrote a new tutorial on SenseCap Watcher, btw.
There is a lot of info. I will try some of this discoveries in Grove Ai Vision V2 itself…