Anyone use LEDs other places around the room?

  • Hi,


    I was wondering, does anyone use lights in other places too around the room, example like Phillips Hue bulbs that synchronize with the TV?


    I think it should not be difficult to do. Just need lots of long wires depending where you want to put but it will need to be correctly placed on the strip(s) for correct lighting colour.


    If anyone has done this, please share.

  • I've got 100 LEDs on my original setup around my 55" Samsung QLED. I later built a couch drink rail around the back and side of my couch in which i in-layed 253 LEDs. I was using the flatbuffer to forward from the tv pi to a pi connected to the couch rail. I then only selected the bottom edge and part of the left edge of the screen for the couch rail to replicated. Its a nice immersive effect.

    https://photos.app.goo.gl/YRyR1dsXdQd7mEbY9

    As a side project I also have a pi analyzing the screen and using computer vision to detect when the Bluejackets (Hockey) score a goal, which i then trigger a custom goal light effect (seen above) on both set of lights.

    • Offizieller Beitrag

    nice one! instead of using a second pi you could use a esp-based wifi controller with the WLED software...


    could you tell us how you implented the CV part? i think more users could be interested with this on :)

  • I've never been super well versed in any of this stuff, but i find solutions to hack my way through it. I do plan on looking into a wifi-controller, but for now the extra pi works ¯\_(ツ)_/¯ . I do want to create full write-up with pics and all someday, just gotta find the time.


    Essentially....

    I started with a service someone else wrote that hit a webserver for NHLStatsAPI every 2-3 seconds checking the score, if the bluejackets score changed I could set a function to happen. I would use the Hyperion remote feature and trigger a custom goal effect i created 8+ years ago. The problem i ran into was that depending on my tv stream, sometimes the lights would go off a few seconds before we scored and sometimes they would go 30 seconds after they resumed play from the goal. I needed to make them locally triggered based on what was on the screen.

    A majority of the Bluejackets games are on Fox Sports Ohio. When they score it would throw a flag in the corner over the scoreboard saying GOAL with a bluejackets logo. I originally went down the path of machine learning and started training models. After some discussions with some people who knew more than me that i found on youtube, i learned I didn't need it to learn anything if i was just looking for an exact match. So i went with OpenCV (computer vision). My only hiccup now was that the flag would change slightly for each type of goal. Short-handed, Game Winning Goal, Power Play Goal, regular Goal. They would also change the logo for what jerseys they were wearing each night and there was also the national broadcasts on NBCSports (only 1 or 2 but i wanted it to cover all).

    I split the RCA signal to the pi (this was already a little dirty causing some slight color shifting) but one continued to the tv pi and a new one went to my trigger pi. I had considered running this all on the tv pi, but the video signal could only be processed from one application and OpenCV on every frame can be processor heavy. There may even be a way to send the frames from the tv pi to the trigger pi via flatbuffer so i wouldn't need the other usb capture device, but again, im hacking here lol.

    OpenCV would take each frame and search for a particular image in each frame and i could adjust the variability rate to account for poor signal. This meant it only had to be an 85% match, i had to be careful though, if i set it too low it would give false positives for opponents scoring and thats not fun. So I screen capped each style of goal and cropped them to just the flags including 2 frames from the NBC goal style (it moved so i had to add an extra instance of it to improve the ability to catch it at my pi's framerate). This accounted for 16 different goal triggers. OpenCV would essentially analyze each frame looking for one of these images. I was able to crop to just the upper left region of the screen to improve frame rate. But with so many images to search for in each frame my FPS was around 4. But it worked!

    That lasted all of 10 games until Fox Sports decided to change to Bally sports and change their whole graphics package (#@&$). But alas it turned out to be a good thing. Even though the flag changed for each type of goal, the same bluejackets logo was used in the same location on the screen every time! This meant i no longer need 16 different images, just the one! My FPS jumped into the 28fps range. As an added bonus as a Reds fan the same regional network carries all the games and they use the same graphics package so i just have to throw a Reds logo in there and it will work for home runs!


    So now once my team scores the networks throw the flag up within 3 or so seconds of the actual goal, i then tell it to play my goal effect on the tvpi and my couch rail pi for 40 seconds (Average time it takes after the goal to get back to dropping the puck for play again)

    I still have to reset the NBC broadcast trigger, but since they are only 1 or 2 games a season, they are not a priority and could even just manual set a separate program for that. They also changed their graphics package so i'll have to wait until the Bluejackets score a goal on their network to capture it.

    I know that wasn't super short, but when i get a better write-up with pictures, i'll be sure to post it to the forums.

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!