![]() |
Parkerj33's Baseball Card Sorting Robot
I took a [URL="https://youtu.be/rQGwanwh9Ho"]video[/URL]... please forgive my crappy videography.
Still have some minor tweaks to do on it....but i have been using it for a week to get caught up on my 2017 heritage base cards before the HHN arrives. It takes about 2 seconds per card....not nearly as fast as i can do it myself, but then again, now i don't have to do anything except load up the stack! EDIT: I should clarify: it sorts the 6 inch stack by hundreds first. then each stack of hundreds are run through the machine again to be sorted into 10s. More details: I started with a stock robot called the uarm Metal. ($330 online). It did not come with a camera. I work for a company writing software that does machine vision... i used one of our cameras and wrote custom software to detect the numbers on the cards. I had someone I know print the holder and stand for the robot using a 3d printer. I had to write software for the robot, and then wire up some custom cable to connect the camera to the robot. I then further modified software on the camera to output the information to the uArm controller so it would take that info and know what stack to put the card in. The hardest part (by far) was getting the robot arm to move properly to gently pickup the cards. Probably needed a more expensive robot, but it does get the job done. |
AMAZING!!!!!
Dont even have to be there for it to sort! |
Really awesome robot.
Thanks for sharing! |
Genius machine.
Get your costs down and/or speed it up, sell it to big breakers. |
If you speed it up I would be interested. I despise sorting so if this could take it into piles 100's and then down to 10's it would be a major time saver for me.
|
That's damn cool! Thanks for sharing.
|
This blew my mind. Color me impressed.
|
Thanks guys. until i can address the static cling phenomenon, i can't speed it up...
|
1. Take it on Shark Tank
2. Make $$$ 3. Buy more cards 4. Put more robots to work sorting said cards Also...Congrats! You accomplished a lot more than I did this week lol |
Good stuff. Very impressive.
I wonder if air blowing in on the sides of the cards would help with the static cling. |
Sweet.
|
that's awesome, but i can't replace my wife with a robot for card sorting she would not be happy
|
- Rub it with Bounce sheets
- I like how the left two stacks are former Brewers |
air blowing might work....i have to wait up to 2 seconds for the cards to fall off....thats why the stack holder has corner guides, so when the cards fall off they gently fall back onto the stack.
|
I don't know what to say. That's incredible!
|
added more details to first post.
|
Meh, does it sort light backs?
|
Haha, that's fantastic. I actually started writing some software dealing with image recognition for cards but got in a little over my head and got impatient (I'm a software developer) - this makes me want to try it out again. You did some really awesome stuff!
|
[QUOTE=jewcer2k5;12718371]If you speed it up I would be interested. I despise sorting so if this could take it into piles 100's and then down to 10's it would be a major time saver for me.[/QUOTE]
This is a spectacular start! So while Jim gets this going full-speed building sets and making these at mass production for us all to affordably own... in the meantime...have kids or hire kids will be quicker and cheaper in short term while waiting |
Very Impressed. You may be onto something. As Jewcer said, I am very interested if there is a way to speed up the process slightly.
|
Real world problems
|
The start of machines becoming self aware.
|
No need to produce a machine to do a job that a mother in law could do
For free |
This is pretty cool.
You could eliminate the static cling by re configuring the angle of "the stack". The stack of cards shouldn't be perfectly flat, it should be at a 45 degree angle, just like a deck shoe (the deck holder that a dealer uses for blackjack) and the cards should be pulled off to the side, not straight up. Assuming you have no static cling, that speeds up the machine from 10 cards per minute to 15 cards per minute. Can the arm move faster than it is now? Also, does each set need it's own program or does the camera perform zonal OCR? |
[QUOTE=Master Shake;12718914]The start of machines becoming self aware.[/QUOTE]
As long as we DON'T turn skynet on... we're okay. |
Awesome start! Team sort would be the next frontier!
|
We have the technology.
|
That is awesome!!
|
Team sort is tough because this robot does not have to range to create 30 different piles of cards
Sent from my iPhone using Tapatalk |
[QUOTE=parkerj33;12719342]Team sort is tough because this robot does not have to range to create 30 different piles of cards
Sent from my iPhone using Tapatalk[/QUOTE] MORE ARMS. :)! |
[QUOTE=parkerj33;12719342]Team sort is tough because this robot does not have to range to create 30 different piles of cards
Sent from my iPhone using Tapatalk[/QUOTE] Subsort like 100's then 10's. Seems like it has to have programming to recognize the numbers, do the same with teams and just bin them maybe by division (if team = "Red Sox", "Yankees", "Orioles", "Blue Jays", "Rays" then AL East pile, etc). Also, as for the speed, even if it is slower, you could always be sorting a pile while it sorts...that or you need 2-3 robots... |
Yes. Very easy. But would need to bin several together and then resort. Or do the final automatically.
Sent from my iPhone using Tapatalk |
WOW! This could be the end of the nightmare mountain of base cards in my basement and the 6 hour marathons of looking for the "hot" player.
My one tweak though would be a stack holder that contains the cards by the edges, not the corners, like graded card cases to reduce the chance of corner wear. |
I always find it funny when people who don't know how hard to do something like this comments with a "meh".
For the first working prototype this is a very good start. Also the key thing here is you put the cards in and walk away or do it overnight. As far as the static thing I deal with that in my lab when weighing up powders. They make anti-static plastic and static wands for use in labs, you might want to see if you can get some to try out. |
[QUOTE=Morgoth;12719498]in my lab when weighing up powders[/QUOTE]
Definitely nothing nefarious... isn't that true WALTER WHITE! I kid, as you are absolutely correct. As someone who writes software for a living I often hear "can't you just whip up an app?" or "can't you just build a new website that does XYZ?" like it ain't no thang. |
Parker, this is incredible, really. When I've laid out designs in my head (I thought of doing something like this in 2012), this isn't at all what I pictured.
Really got my brain working this morning. :) |
Most impressive!
|
This is awesome! I so need one of these haha
|
Very impressive, well done!
|
Very nice!! I had to make something like this in school once, but it sorted like gumballs by color or something stupid like that. This is much better.
|
That's really amazing! Nice job.
Sent from my SM-G930V using Tapatalk |
1. Robots
2. ??? 3. Profit |
[QUOTE=tsnider45;12719703]1. Robots
2. ??? 3. Profit[/QUOTE] YES. Also, incredible user name. Snides! |
Thanks for all the praise...and also for the inspirations for a few improvements. Ver2.0 on the way....
|
[QUOTE=mattglet;12719513]Definitely nothing nefarious... isn't that true WALTER WHITE!
I kid, as you are absolutely correct. As someone who writes software for a living I often hear "can't you just whip up an app?" or "can't you just build a new website that does XYZ?" like it ain't no thang.[/QUOTE] At one point, I could write code in 5 languages (probably none relevent today). I understand that sometimes things that may seem easy can be the most frustrating to implement properly. Changing sorting from a range of numbers to specific words could be a pain. I do think this is really cool. |
This is awesome - I teach robotics at a middle school and love showing students this kind of stuff. Let me know if you would be okay with that.
|
Please show away!
Sent from my iPhone using Tapatalk |
That's so damn cool
|
Thanks!
[QUOTE=parkerj33;12720278]Please show away! Sent from my iPhone using Tapatalk[/QUOTE] |
Just forwarded this to the wife and told her she needed to step her game up......
|
This is awesome, better than I expected when you mentioned it yesterday!
I do have a question, how did you get it to read the puzzle piece back correctly? If I recall, the number for those is on the bottom, so I'm assuming it would have to scan the entire card for numbers? How do you differentiate between the card number and the other numbers on a regular card in that case? |
ahh...you have found a flaw! i did not bother to read those....in the interest of getting it working quicker, i didn't bother to add an algorithm to detect those numbers...they are in a completely different location and in the current configuration the camera can only see the top portion of the card.....(the gripper suction cup is in the way of the bottom half).
i have a stack where it puts unknowns when sorting by 100s. so that stack is filled with all the allstar cards (puzzle backs) which still makes them super easy to sort. One other issue: because the team cards have numbers that are rotated 90 degrees from the standard number, i am not decoding those yet either.....again, i didn't mind missing out on those....but these are heritage-related optimizations to make, when my free time allows. To the poster that asked....i am NOT doing OCR...i am doing pattern matching on the numeric shapes....but that is because i haven't yet written OCR software for the camera yet. - jim |
Holy cow that's amazing!
|
[QUOTE=parkerj33;12720651]ahh...you have found a flaw! i did not bother to read those....in the interest of getting it working quicker, i didn't bother to add an algorithm to detect those numbers...they are in a completely different location and in the current configuration the camera can only see the top portion of the card.....(the gripper suction cup is in the way of the bottom half).
i have a stack where it puts unknowns when sorting by 100s. so that stack is filled with all the allstar cards (puzzle backs) which still makes them super easy to sort. One other issue: because the team cards have numbers that are rotated 90 degrees from the standard number, i am not decoding those yet either.....again, i didn't mind missing out on those....but these are heritage-related optimizations to make, when my free time allows. To the poster that asked....i am NOT doing OCR...i am doing pattern matching on the numeric shapes....but that is because i haven't yet written OCR software for the camera yet. - jim[/QUOTE] Jim, you've had me thinking about design ideas for this all day. As I mentioned, your implementation isn't at all what I had in mind. Having said that, your implementation may be [I]better[/I] than what I had in mind, but I'm not sure yet. The only real flaw in your design is speed. You need it to be faster. However, the only way to get there may be much higher cost, and for what you've invested here, the process is fantastic. |
wow this is pretty cool . but on another note if a few armed people from the future show up on your door step. just give them the robotic arm. lol.
|
[QUOTE=calculusdork;12720801]Jim, you've had me thinking about design ideas for this all day. As I mentioned, your implementation isn't at all what I had in mind. Having said that, your implementation may be [I]better[/I] than what I had in mind, but I'm not sure yet. The only real flaw in your design is speed. You need it to be faster. However, the only way to get there may be much higher cost, and for what you've invested here, the process is fantastic.[/QUOTE]
I've always imagined a sorting machine being done conveyor belt style (like a post office letter sorter) or even a system that looks like a clock face. Cards get fed into the 12 o'clock position, then get removed when they reach their sorted position. That way, the "machine" doesn't have to put each card into it's final sorted position before it can pick up the next. THAT, imo, is the flaw of this type of system, beyond the speed of it (don't get me wrong, I love this sorting robot). It's overall limit is how fast it can completely sort an item before it moves on to the next. At best, this type of system can complete 30-35 cards per minute. Again, don't get me wrong, I love this machine, but the sorting method itself is the restriction here and the only real way to "make it faster" is to buy more robots. With a conveyor style, the overall limit is dictated by how fast you can fed cards on to the conveyor. In that case, you're potentially at 110-150 cards per minute. |
[QUOTE=parkerj33;12720651]
To the poster that asked....i am NOT doing OCR...i am doing pattern matching on the numeric shapes....but that is because i haven't yet written OCR software for the camera yet. - jim[/QUOTE] I'm not very familiar with computer vision and processing. Is one version (OCR vs pattern matching) faster than another? In your system, is the pattern matching complete before the arm even starts to move? Because it looks like it takes just under 2 seconds for the image to be processed and for the arm to start moving. Would OCR speed that up? |
the processing is done before the arm raises from the stack. the 2 second delay is not the time to do the vision. that is super fast (~100msecs)....the delay is self-imposed to give the cards time to drop from the clinging.
|
[QUOTE=CharlieHustle;12722897]I've always imagined a sorting machine being done conveyor belt style (like a post office letter sorter) or even a system that looks like a clock face. Cards get fed into the 12 o'clock position, then get removed when they reach their sorted position. That way, the "machine" doesn't have to put each card into it's final sorted position before it can pick up the next. THAT, imo, is the flaw of this type of system, beyond the speed of it (don't get me wrong, I love this sorting robot).
It's overall limit is how fast it can completely sort an item before it moves on to the next. At best, this type of system can complete 30-35 cards per minute. Again, don't get me wrong, I love this machine, but the sorting method itself is the restriction here and the only real way to "make it faster" is to buy more robots. With a conveyor style, the overall limit is dictated by how fast you can fed cards on to the conveyor. In that case, you're potentially at 110-150 cards per minute.[/QUOTE] Yes. Conveyor is what I had imagined. Card feeds in, image taken, then some sort of filtering system on the conveyor to move the card to its proper stack/location. Camera feeds information to the filtering system but is not inhibited by it in any way. I imagine this is a simplified version of what COMC uses. Jim, thanks for getting me thinking about this. :) |
yeah conveyors and movement systems is way beyond what i care to consider....thats industrial-caliber.
|
[QUOTE=tsnider45;12719703]1. Robots
2. ??? 3. Profit[/QUOTE] 2. Patent 2b. Licensing deal |
[QUOTE=parkerj33;12723256]yeah conveyors and movement systems is way beyond what i care to consider....thats industrial-caliber.[/QUOTE]
Indeed. But it's not going to keep me from thinking about it way too much. :) |
Very impressive, thanks for sharing!
Would it be possible to have the robot search the small print for SP and SSP's in various Topps products including Heritage? So go through the stacks and pull out all the Heritage actions, errors, variations first? Then sort by card number? |
[QUOTE=Coug46n2;12723298]Very impressive, thanks for sharing!
Would it be possible to have the robot search the small print for SP and SSP's in various Topps products including Heritage? So go through the stacks and pull out all the Heritage actions, errors, variations first? Then sort by card number?[/QUOTE] I think all Jim would have to do is capture the image of the fine-print code at the bottom, and use the same logic to differentiate. |
[QUOTE=calculusdork;12723211]Yes. Conveyor is what I had imagined. Card feeds in, image taken, then some sort of filtering system on the conveyor to move the card to its proper stack/location. Camera feeds information to the filtering system but is not inhibited by it in any way. I imagine this is a simplified version of what COMC uses.
Jim, thanks for getting me thinking about this. :)[/QUOTE] Jim - this is awesome! Grant - I bought all the stuff and had this what you are talking about started, but never finished. Stepper motors, raspberry pi, camera with OCR python program, etc.. Still in boxes, lol! Once it transfers to the conveyor was going to start the next card in the process. this work here makes me want to go finish it. |
[QUOTE=mjohnatgt;12723265]2. Patent
2b. Licensing deal[/QUOTE] unfortunately, #2 becomes much harder with showing that video unless already filed with USPTO. |
Smart, creative people are awesome. Nice work OP.
|
Just an update...I have successfully sorted well over 25,000 cards with the robot (mostly heritage but also 2018 topps series 1).
Some noted observations: 1. The hopper now holds about 400 cards and empties in about 20-25 minutes with about 10% unable to sort. (3-4 seconds/card). 2. The robot is not perfectly accurate in its placement so the piles get a little messy. i have to clear them out after about 100 cards (per pile). This is a limitation of the robot i am using. Will likely upgrade to a pro version of the robot ($800 vs $400) which uses much more accurate stepper motors, not these sloppy servo motors. Will also be much quieter. 3. Machine vision has needed tweaks to be more accurate. Very few mis-reads (<1%) but I still get about 10% failures (unable to read ANY number). This is still too high for my tastes. 4. I have improved the speed significantly over the video...but still not down to the target of 1 second per card. 5. Modified the carrier so that there are only guide posts that are in the center of each side of the card, not the corners. 6. The machine vision camera has to be taught anew for each set i sort (2005 heritage, 2008 heritage, 2017, etc. It only holds the programming for one set, so i have to load the right program in per set. But i only have to train the camera once. |
Keep the updates coming Jim - it's still awesome!
|
Incredible. That’s impressive!
|
I want one.
What equipment does/did Topps or Donruss use? JOBO's are still the rage for Black and White film photography but they are becoming quite the treasure as supplies dwindle. There has got to be some old technology floating around to rehab or copy. |
[QUOTE=chfu3030;12720403]Just forwarded this to the wife and told her she needed to step her game up......[/QUOTE]
Ywoe, that could get ya hurt:eek: |
I wonder if it would be possible to have it read the code on the back of the card? That way it could pull out and sort inserts, parallels, and SP. Although I guess it's not that hard to manually pull them out while you break but it would still be cool.
|
Wow Jim, I am impressed , good luck on the 2018's !
|
[QUOTE=parkerj33;13284268] I have to clear them out after about 100 cards (per pile). [/QUOTE]
couldn't you just set up some simple supports around where the piles are deposited, say just simple blocks of wood 4" deep by 3" wide? |
| All times are GMT -5. The time now is 07:58 PM. |
Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2025, vBulletin Solutions Inc.
Copyright © 2019, Blowout Cards Inc.