This is an interesting idea, and we agree that preorders generally can be problematic since there is so much uncertainty in product manufacturing. It really hurts to push the launch date back, but I think it is the right thing to do in the end. We opened up orders yesterday because we had fully tested assembly line units in-hand and began shipping to our distributors.
2 Likes
We have ours from last year. Is there a planned release for 2020 firmware?
Latest I see on the sight is 2019.7.1.
1 Like
Yes, 2020.0 will be posted to the website today, and 2020.1 with very cool new features will release soon.
1 Like
Thanks for the update Brandon. No 2020 image yesterday, today maybe?
1 Like
The new downloads page just went live with the 2020.0 image as well as the Finder Tool. New PNP models coming tonight. (And new 2020 features are still coming soon in 2020.1)
Such a tease!
1 Like
Hey everyone, we have had some trouble with the new 3D features of our limelight 2. We have successfully got the thresholding working and managed to filter out everything except our target and now we are trying to get the robot to detect the shape of the target. There are no 2020 targets on the limelight website so we made our own (pictured below) For some reason, the limelight will not detect the shape of the target and generate the vectors as shown in the patch notes here. We have checked the units of the spreadsheet and our image is pretty well processed(we have filtered out all the other lights from the image and we have minimal flickering with a relatively sharp image). Has anyone else had a problem like this and if you have how did you solve it? We are going to try processing the image a bit better and I’ll leave an update if it works.
Edit: I just saw that the 2020 PnP images are coming out tonight. We will also try using those instead of our homemade ones to see if it works
They were in stock effective at noon today, but were out of stock by 6pm (likely earlier). Last I checked, they’re still in stock on the limelight website though.
Update to the above post, we have tried using the new pnp images and we still have had no success with the compute 3D function of the limelight. We have very good thresholding but the 3D function still will not detect the shape and give us any information on our position to the target. Does anyone have a fix for this?
1 Like
Just sent you a PM. The new image is likely releasing by the end of the day.
I had the same problem yesterday with LL1. Will this new image fix on v1 too? Or if there’s a separate v1 image, will there be an update to fix it for v1?
It will fix the V1, V2, and V2+!
2 Likes
Here are the new release notes with some new solvepnp features -
Is there any documentation on how to get started with the 3D / Solve PnP functionality. I understand the theory but I can find any Limelight “He’s how you get started” docs.
A new documentation overhaul is going to start manifesting at the end of the day, but currently there are no guides for the experimental 3D features. I can summarize it here (assuming 2020.1 image):
- Enable the 720p video mode in the input tab
- Enable 3D solving in the 3D tab (enable force convex for the hex goal if it is not enabled)
- The new image comes preloaded with a csv for the hex target in pipeline 0, so at this point you need to perfect your thresholding to get nice corners.
- Check the nettables API page in the docs for info on grabbing the 6dof transform, or use bind target and use tx and ty for turrets/turning.
Did as you said and I think I have a pretty solid pipeline.
Just doing this initial test my LimeLight was a good distance from the target. 25 feet or so.
To me the targeting was great. Solid, almost non moving rectangle. I would be confident targeting using the info from the LL.
But, I turned on 3D and it showed the points on the camera image, 0,1,2,3 but noting in the 3D view. On occasion I would see some activity for about half a second.
In the camera output on the webpage the numbers in the upper left were 22.0 and below a zero.
Have no idea what those numbers are but if I changed camera to 2x zoom the top number changed to 90.
Do I just need to test closer?
From our experience, the 3D feature is not reliable at distances greater than 7 feet or so. It requires the edges to be crystal clear.
I also don’t believe you are using the target model for this year’s game, you can download the 2020 Hex Goal model from their website and upload it to your pipeline.
You’ve run into a corner correspondence issue that we are going to address in the next update. In general, we are attempting really improve solvepnp stability at long distances and during movement with some of our own new solutions before removing the “experimental” tag.
1 Like
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.