I don’t have any solid numbers, but when using a coral with the LL3 I got around 30fps (from what I remember, not sure if this is 100% true or backed, would love if someone has a more solid comparison)
Even when testing with the same model on both we got better results with the fork.
I’m really sorry and I don’t want to confuse or annoy anyone. I truly don’t. This is me posting, partly on the behalf of Lavi with his information (since I don’t know all the technical parts of this project)
I don’t know if I’m stepping into a controversy, all I want is to just make this public to all teams. If people get annoyed with me, even though I truly don’t mean to, then so be it
Either way I really don’t mean to hurt anyone and I just want to be a part of the FIRST community as much as I can
This isn’t supposed to be some kind of big server, just a small one for testing that whoever wants to help can join.
I don’t even know what to say about this lol. What? He just created a discord server, he’s not DM-ing minors or harassing anyone.
This code doesn’t modify any of the JNI libraries, and was almost completely written before PV started working on their implementation.
That is interesting. Guess I’ll have to spend some time digging into the differences.
(thank you auto complete for suggesting a thumbs down… I woulda felt real bad about that)
OK, it was a little unclear so I wanted to be sure. Lavi has demonstrated they were not the sort of person I wished to engage with.
I will say, who you post on behalf of and who you engage with do reflect on you. But as far as I can tell you’ve been doing a great job of contributing to the community. Bluntly, even if this account was simply a Lavi I’d be more than happy to assume they’d learned how to more positively engage in this community. So, no worries from me, you seem to be doing a great job.
Unfortunately, the association does force a lot of us to raise questions - I’d be real cautious of encouraging students to engage in any community with Lavi that wasn’t 100% public and had at least made efforts to protect students. It all really sucks that a) you’re caught in the middle and b) that someone as, obviously technically talented as Lavi is unable to engage directly for whatever reason.
Mumble mumble discord in general mumble mumble poor data retention practices.
I really wish this could be done on some sort of more public server with more protections, the proliferation of small servers within this community scares the carp out of me. But this is way out of scope…
Anyway, TSIMFD and I really encourage you to learn as much about it as you can. It’s cool. Just, you know, don’t take advice on how to engage with a community from someone banned from multiple of them I’ve been on CD nearly 2 decades at this point and only been suspended once for a week. And ask most people, I’ve not always been the most welcoming or encouraging person.
All of the features are supported in the current version of photonvision. I’m not sure if optimizations are in the current release of photonvision, but uploading and selecting custom models are both currently supported, with v5 and v8.
I would also recommend using the official version of photonvision, as that’s going to have more support, and be more recent. This fork looks to be out of date by about 5 months, unless I’m looking at the wrong repository.
Although I highly dislike the way PV implemented the UI to upload models, most of the features are there.
Mainstream PV seems to be kinda half broken for all pipelines, and for RKNN specifically, non quantized v8 models don’t seem to work properly, and also not utilizing the NPU properly.
In spite of the above, I don’t have the time to maintain the fork, so the official implementation would have to do for now.
(thank you auto complete for suggesting a thumbs down… I woulda felt real bad about that)
I’ve been on