We are looking to buy a google coral but can’t find stock anywhere. Does anyone have any idea where we could buy it?
Andymark recently started selling the Google Coral USB Accelerator. This is convenient if you have credits or vouchers to use on Andymark as well.
Check it out here.
5 Likes
Amazon surprisingly always seems to have one in stock. They are SO worth it, I HIGHLY recommend them. Especially for teams who dont have the time for april tags and or want more consistent tracking.
Although I agree that the Coral is a valuable coprocessor for FRC, it is only useful for accelerating machine learning vision models. It is not a generic accelerator for OpenCV processes. Teams that “don’t have time for” april tag tracking almost certainly don’t have the time for developing a machine learning vision model. And if they are using an off-the-shelf vision solution then tracking april tags is no more complex than tracking game pieces using a ML model.
I agree with your overall sentiment, but your comment could leave teams with overhyped expectations of what the Coral can do for them.
9 Likes
Interesting perspective. In my eyes, the reason why the coral is so great is that in conjugation with the limelight, It’s pretty much plug and play. Every year you can find multiple premade models of the game objects for the limelight, even official ones by the limelight company. This means that you plug the thing in, do some downloading, and then you’re straight into programming. With April Tags the best way to use them is with path planner which is a headache in its own right, and really I think that’s the kicker. I know a team who does basic “move for X time” and “rotate X degrees”. The Google coral has really helped them pick up game pieces at all without using path planning, which in my experience can take far longer to figure out.
If you hold a different perspective, I invite you to share it, as this has simply been my experience as a programmer with it.
I appreciate your concern for the larger community, and I do probably have a little extra excitement in my voice as it’s as fun new toy our team gets to play with, which is something I should better consider when writing these posts. However, I do believe that these devices still hold tons of value for teams that don’t have the time for path planner and April tags.
There are much better ways to run object detection on a robot.
If you are already using a LimeLight then you can already get april tag localization “for free” without adding an accelerator or downloading any models that may or may not exist during competition season.
From that point, the Network Table data are in the same basic format, so coding knowing that “I see a game piece here” is the same complexity as coding knowing “I see an april tag here.”
All of that is before you start to integrate the april tag location data into your odometry. And all of that is AGAIN before you start using Path Planner or some other trajectory planning tool.
To repeat, I agree with your enthusiasm for the cool things you can do if you add a Coral to your vision system. But in your enthusiasm for the gadget, you failed to mention any context for how it might be useful, while using language that implied that it would somehow provide “more consistent tracking.”
If your premise is that a team lacks the time or knowledge to use the free tools already at their disposal, then advising them to buy a gadget that solves a very specific problem should always come with that added context.
The way your original post was worded implied (to the casual reader) that a Coral would improve tracking of april tags, or that it could be incorporated without any coding effort. Neither of those things is true.
2 Likes
oh 100% but this is just so easy Ig? Idk
100% and im not saying that teams should do this over april tags all the time, in fact for most teams april tags are far better, however with autos, april tags are really only helpful if you also have path planning, which can be a hassle. All im saying is that it can be technically easier to understand for some teams sometimes.
Yes, I agree I didnt give proper explanation for my claims, nor did I properly give context, however, the goal of that post was not to inform, nor to make a claim, but to simply share my enthusiasm for the product and encourage the OP to go through with purchasing one. I am not an expert on this, and I was simply speaking from my own experience. In fact I appreciated the context you added, however I am simply defending my original claims as I believe them to still be valid.
You are correct, except that I was not encouraging them to “buy” it I was simply sharing my enthusiasm for it. A random post about an individual should never be taken as fact, and they should always do research on their own. From my perspective, I simply was encouraging teams to look into a tool that could be helpful, not making wild claims.
I stand by that for the use of autos this can be a viable option for some teams over April tags. I additionally stand by that it can give more tracking options of game pieces when used in conjunction with april tags and therefore give you “better tracking”
I am no English major, nor am I the best at giving context. I appreciate the added context to the original post, and I’m simply working to amend it. Though, I am confused as to why it is so important.
2 Likes
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.