hi there. i’m from an frc team and i want to know whether photonvision or g rip or limelight is bettter? limelight is around since 2011 and grip and photon are newer, so is the newer stuff bettter than the old stuff or is the old stuff better than the newer stuff? i ask because i dont know if i want to spend the 400 on a limelight if grip is newer and better. we are smaller so the extra 400 is not much but is not too little, again is it bettter for the money?
4 Likes
“Better” is a relative term. Properly applied, there are a lot of solutions which get you to the same end functionality.
Limelight is off-the-shelf, end-to-end, hardware and software included. You put down $400, do a tiny bit of tuning, and end up with a vision system that tells you various pieces of information about your distance to the target.
Photon Vision is newer and provides a similar software functionality, but you have to source your own hardware. Gloworm is in beta and provides a fairly off-the-shelf solution. A raspberry pi and camera and case and LED ring will get you close. It’s cheaper, but it’s more hands on.
GRIP is a nifty tool for learning, but won’t solve the problem of “vision processing” on its own. You can use it to deploy to various targets, but few teams were using it (and WPI hasn’t updated it in a while). It will require some level of custom software work to integrate into any real-time hardware platform, adding another level of complexity. However, it offers more flexibility in defining a complex vision processing pipeline over the other two options.
My recommendation: Try to get something set up on a raspberry pi this fall with Photon Vision, and see where it gets you. If you don’t have something satisfactory by December, drop the $400 on a limelight.
9 Likes
Alternatively drop ~$150 (shipping handling) on a Gloworm which is near identical to a LL, easy imaging with PhotonVision, and works just as good.
To be honest, unless your team is fairly experienced with Linux and programming in general, I wouldn’t recommend jumping straight into Gloworm/PhotonVision at the moment… Gloworm is in beta and (more importantly) so is PhotonVision. Like @gerthworm said, you should try out PhotonVision, WPILIB’s FRCVision, OpenSIght, etc, on a regular Raspberry Pi in the offseason, and if you get far with that then consider a Gloworm. Otherwise, you’re probably better off with a Limelight for now.
7 Likes
Have these been run through a competition season with a variety of teams?
1 Like
PhotonVision at least has in a way, as it’s a fork of ChameleonVision. It doesn’t look like people had a ton to say about ChameleonVision either way, though, so take that as you will.
No, but if the OP is complaining about $400 (and are a smaller team), then they’re more likely to chance it.
2 Likes
First of all, it is important to let you know that PhotonVision and Gloworm are both in their beta testing phases at the moment, with their official releases coming soon ™. PhotonVision still has many bugs in the software that will hopefully be mainly gone after this beta testing phase.
If you need something now, the Limelight is a great, tried and tested vision system that is almost guaranteed to work for you. It requires minimal setup and there are comprehensive docs, videos, examples, and a huge FRC support community ready to solve your problems.
GRIP is rather old and is what some would classify as abandonware. It helps you learn the what/how behind a lot of vision processing but is going to be a lot of unnecessary work to set up, especially as a team not knowing much about vision processing. Yes, it allows more flexibility but it really depends on what you are doing.
PhotonVision is new and a fork from the ChameleonVision project. It has a new UI, performance updates, etc. and possibly has benefits of a LL in terms of software performance. However, it is not a polished product yet and still needs some work for it to be ready to be used by the general community. Documentation is also not ready. You can use existing hardware you have or buy a Gloworm as users have talked about above. This is significantly cheaper than a LL but it is not as tried and tested. It is getting there, but it needs more time and a season to put it through its tests.
Basically, I agree with @gerthworm and @fharding completely but didn’t wanna throw away this post.
At the end of the day, it is up to your team to evaluate your abilities and what you are willing to invest more money in or to take a risk on.
Disclosure: I am a really inactive member of the PhotonVision team.
1 Like
A small team is often the last team that should be taking a chance on a solution that may not work correctly. Gloworm / Photon / etc may all work a treat once you get them up and going. But as a team with highly limited resources, what the original poster should be asking themselves is:
- Is $400 out of my price range
- How much time am I willing to spend working on the vision system of this robot versus addressing other parts of this game challenge.
If they want a drop-in solution that lets them move on to other portions of the game, the limelight is probably the best answer. If they want a solution that may or may not work without a substantial input of effort, then the others may be potential solutions.
The OP needs to better define their need, and do a decision matrix to help their decision. A question the OP might ask themselves is: Am I confident I’m going to have a shooter that is repeatable and robust enough that the limelight is even needed? Or can I use a flashlight for aiming and put my resources elsewhere?
2 Likes
Just to be clear, Gloworm + PhotonVision is intended as drop-in solutions with no extra work. The amount of steps to setup a Gloworm with PhotonVision is near identical to setting up a limelight (download an image, flash, and go). The only risk involved is stability, as Gloworm and PhotonVision is WIP and beta software.
Gloworm is the name for the hardware package itself (think of it as the limelight hardware) and PhotonVision is the software (think of it as the limelight software). Additionally, PhotonVision has feature parity with LL alongside some additional features. However, again, there are bugs so there is a risk.
Just wanted to make it clear that Gloworm and PhotonVision are intended as plug-in-play.
2 Likes
And I bet it’s going to get there, especially with all the progress we’ve been seeing here on Chief. However - until it’s been run by numerous teams in a competition environment over a span of time, we won’t know if it lives up to its promise.
That’s not a negative, it’s just a rule of nature. It’s the same reason people steer clear of a car the first year it’s in production. Until there are some miles under its proverbial belt, its too soon to steer users to it as a solution that is comparable to a limelight.
Frankly I’m looking forward to it. Every year I get asked if we really need to use the limelight because of its cost.
FWIW I started playing with the beta this weekend, and was thoroughly impressed. I think it’s got great potential to be the main way of the future for folks who want cheaper or more flexible options than Limelight. Still, in its current state it’s got some issues. Namely, I had to know ssh/systemctl/wget to get the latest software running. USB camera support was varied and isn’t fully polished yet. Some key documentation is still largely missing. A lot of this will likely be fixed prior to the season, but I don’t know for sure if that’s a guarantee.
It’s got excellent potential, but still needs some polishing before I’ll sign up to say “it’s actually a Limelight replacement”. IMO in its present state, your success will vary depending on how far you stray from the known/“blessed” path.
Still, Limelight was indeed revolutionary for vision processing, so saying an open-source/community-developed software/hardware suite can match it is a very high bar to hit! The fact the team is close is something to be proud of.
Hence my previous recommendation: For any team with time now - start the investigation to see whether you can make the cheaper/open-source solution function with all your team’s assumptions of hardware, vision processing pipelines, linux abilities, etc. SIgn up for their Beta program, grab the latest development you can get your hands on, and stick with it as much as you can. If you hit the end of the year and don’t have something stable enough to trust it as a drop-in solution during the build season - buy the limelight. Multiple years of teams have proven it is indeed a drop-in solution.
is limeliight worth 2500 more than the glowworm? 400 might be a lot for us but i think we can savncege 150. i don see what youre saying though and because it says here that glowworm is drop in but im not sure if its worth the 150 over grip which is free. i like free stuff but is there anything to care about before i start trying?
GRIP is free, but the hardware needed to run it isn’t. With the cost of a Raspberry Pi, an SD card, a camera, an LED ring, and a good 5v regulator in mind, the Gloworm isn’t too bad of a deal when you consider all of that.
Yea, same as Bobby - I’d not think of GRIP the same way as you think of limelight or gloworm+PhotonVision.
GRIP is a tool to help you develop algorithms for processing images, and learn about what sorts of things work (and what don’t). It can generate some code for you, but doesn’t really attempt to dictate what hardware you put that code onto. This is a double-edged sword - you can have some good flexibility, but have to have some pretty advanced knowledge of openCV and whatever hardware platform and cameras you’re using for your vision processing.
Gloworm+PhotonVision and Limelight are full packaged software+hardware solutions to actually do vision processing. Limelight requires nearly zero knowledge of the hardware and camera. Gloworm+PhotonVision require a bit more, but not much (and are requiring less and less as time goes on). They’re more of an “end-to-end” solution, rather than a single tool in a toolbox.
1 Like
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.