Piqabrick - LEGO identification system

[Stefano Berti]

Thanks to Artificial Intelligence, Piqabrick instantly identifies Lego bricks’ code, simplifying the long-lasting operations of searching and sorting.

Among the main needs of AFOL (Adult Fan Of Lego) community, identifying single bricks (maybe rare or unique printed bricks or minifigs) is certainly on the top of the list. But also identify two similar bricks, or discover their related codes could be complicated, since they’re hidden or difficult to read. Not to mention the stressful research on online catalogues or marketplaces!

In a split second, and with just one photo, Piqabrick identifies any brick. 

Sort less and play more.
Piqabrick easily and quickly identifies any Lego brick providing you the ID code and color code. How? Thanks to our proprietary computer vision technology. Piqabrick “looks at” a brick to identify it, just like we already do.. but better!

PIQABRICK relies on DART (Direct Acquisition and ReTrieval), Getcoo’s proprietary Artificial Intelligence, already adopted in industry and tourism. PIQABRICK is made of two parts: the PIQABRICK BOX and the web dashboard. The first one consists of a USB camera and LED (to provide the correct illumination) mounted on top of a box made by LEGO® bricks. The box has precise dimensions (16x16 brick units wide and 12 brick units tall), but it is fully customizable by the user with his/hers own bricks (as long the internal dimension are kept, the color of the bricks does not matter). The PIQABRICK BOX is connected to a PC via USB and does not require any installation. To identify a brick, down to its ID and color code, just put it into the box!
The web dashboard controls the camera in the PIQABRICK BOX for the brick identification. Once the brick is identified the dashboard provides the links to the LEGO marketplaces to easily buy/sell the brick and to manage the personal inventory.

With PIQABRICK, Lego lovers can save time in the brick identification and the personal inventory management. The computer vision speeds up the buying/selling activity on the specialized marketplaces. With less time needed for searching, sorting and inventorying, more time to play, build, and have fun!

Functionalities:

• identifies any brick down to the ID and color code
• speeds up the identification with computer vision
• identifies minifigs as well
• suggests similar bricks (by code or color)
• provides “appear in” functionality (list of sets in which a brick appears)
• provides “rebrick” functionality (list of sets one can build with the bricks in his/her inventory)

More info:

https://pickabrick.live/kickstarter
box.piqabrick.com
www.facebook.com/Piqabrick/
https://youtu.be/Umub6GvWVqk

[Pix]

It's amazing!

Edited October 28, 2019 by Pix

[GianCann]

Some test, with my prototype (I'm one of a ten selected beta-tester)

If you want to see the full unboxing of the product and other picture of scanning system, take a look here

Edited November 2, 2019 by GianCann

[Toastie]

Errr - so "I" (as in me) do see a 2x2 tile with "LEGO" pattern on it - for me - a 100% match.

When not knowing the part, I take my magnifying glass, look at the number type that into Google with "LEGO" and "part" as additional search items - and - almost 100% match.

I am just saying that when hard/software can do it, homo sapiens can do it, right? I simply don't get the idea. Are there too many bricks out there that "we" can't recognize? On the other hand, how does the software do that?

I am lost ... it may also be time - maybe time is an issue here. Is that it?

Best
Thorsten

 

[Stefano Berti]

In case you haven't seen Piqabrick in action yet, here is a review by one of our beta-tester Simon from Germany
https://youtu.be/DZaSa4Zezsw

 

[MAB]

What assurances are you making that the software will be kept updated with new parts and available over, for example, the next 5 or 10 years? You only need to look at another LEGO kickstarter - sbrick plus - to see problems when they don't properly serve paying customers.

Also, do you have an agreement with bricklink to be able to use their numbering system commercially? If that got removed, then your device would lose a lot of its usefulness.

You also say "With less time needed for searching, sorting and inventorying, more time to play, build, and have fun! " How does this save time with inventorying at bricklink, for example? Does your software interface with bricklink, and automatically change inventory levels at BL?

 

 

[Toastie]

@MAB

Did not even think about that - but: Very good points.

I'd like to challenge again though: What is the thing here (because I like very much what you accomplished, but that is not it: You want to sell something. And that changes everything):

• Time saving (against common sense + magnifying glass + google or BL): I have a 1x2 tile; printed on it is "E&L", two hearts, and "4EVA" - Google finds that in 0.00000001 s
• Not wanting to do that
• Being of help for let's call them rookies? I for sure can tell a thick from another type of clip. Plus: The number tells me that.

So when it comes to sell things (as opposed to making it work), why would I? Just tell me - and I will.

All the best
Thorsten 

[knotian]

'Bout the only thing I can see this useful for is parts without any id number - and there are plenty.

[MAB]

When sorting mixed lots, I often find most problems are due to non-lego parts. What happens if you put a clone 2x2 brick in the device? What about a non-lego accessory that looks a bit like lego but not quite?

[Stefano Berti]

5 hours ago, MAB said:

When sorting mixed lots, I often find most problems are due to non-lego parts. What happens if you put a clone 2x2 brick in the device? What about a non-lego accessory that looks a bit like lego but not quite?

@MAB Piqabrick will not detect fake Lego in case they are clearly non-Lego. We discussed yesterday at Bricks in Florence Festival about the possibility to identify Lego fake parts as a fake revealer. We put this functionality in the roadmap. 

[spzero]

I would love to know about mismatched minifigs, say I put an Iron man Mark 7 armour figure in but he has the wrong legs maybe iron man mark 42 legs what happens?

Is it unidentifiable?

Will it show as the Iron man Mark 7 and say legs are incorrect?

Will it identify all parts present? and show that I have head helmet torso of iron man mark 7 but legs of iron man mark 42?

or do I need to identify each part separately

Thanks I am a backer and hoping it makes funding

 

[Stefano Berti]

@spzero We discussed about this functionality with a lot of AFOLs at Bricks in Florence Festival last weekend and this is a real need for a lot of AFOL, so we decided to add this functionality to Piqabrick. It will identify the whole minifig and all of its parts. So you will be able to detect a mismatched minifig. Of course you can also put each part in the Piqabrick box for individual identification as well

[Toastie]

How much faster do you judge your artificial intelligence powered system is as compared to natural intelligence having access to BL and all other LEGO parts databases as well as Google?

I am just trying to figure out how and under what conditions to use this device productively - the parts have to be put one by one into the device, right?

A number of rare parts within a big chunk of plain vanilla parts cannot be the scenario, as the big chunk wold easily be categorized by human intelligence (studs with LEGO printed on them etc pp.) as such. The remaining rare parts then need to be categorized as well. But then there are these categories, which are used naturally by natural intelligence: Brick, tile, slope, minifig part etc. etc. Then maybe pattern, print, color, then Google or BL or whatever. You can find the 1x2 tile in Emmet's Dream house set with the "E&L 4 EVA" print in zero time as 1st entry on Google. OK that one is no way rare. But so far it always worked for me - even with my 1965 train set. I found the parts among many others in the attic of my parent's house a couple of years ago and had them sorted rather swiftly using - Google.  

Where exactly is the benefit?

Sorry, but I really can't figure it out - but that does not mean anything. Just enlighten me (oops - hopefully the EB AI will not catch that one and replace it with "megablocks"

Best
Thorsten

[Philo]

8 hours ago, Toastie said:

How much faster do you judge your artificial intelligence powered system is as compared to natural intelligence having access to BL and all other LEGO parts databases as well as Google?

AI can do wonder from a simple photo... I often use Plantnet  app to identify plants I see during my walks, and I am always amazed by its performance! Now I guess this is less useful for an expert!

[Calabar]

Hi and congratulations for your creation.

At start I hoped it was a simple app: with a limited cost (maybe a free app ) I could have access to this useful identification system, but this solution is too expensive for the very limited use I can do of it. Anyway I can understand the reasons behind your choice, even if a less precise app version (maybe requiring a bit of manual tuning) would be desirable for me.

I read here some users that doubts about the usefulness of this solution. Sometimes I needed to identify some parts from a mixed lot (usually in order to detect from some specific bricks the sets from which the parts came from and try to rebuild them) and this operation was very time consuming. Other times I need to place the right face on a minifigure (or simply understand to which minifigure that spared face belongs), and manually identify a face is often a pain.
So, for my experience, it would be very useful.
For those that says that manual identification is as fast as the one provided by Piqabrick I can only think that these people have above-average sorting skill or never try to do this and make a wrong assessment of the real difficulties and time needed.

[Stefano Berti]

Piqabrick has landed on Indiegogo.
Now you can pre-order it with a special discount on the retail price.

https://igg.me/at/piqabrick/x/22722823#/

[Toastie]

On 11/13/2019 at 4:43 PM, Calabar said:

For those that says that manual identification is as fast as the one provided by Piqabrick I can only think that these people have above-average sorting skill or never try to do this and make a wrong assessment of the real difficulties and time needed.

Ha! Finally someone admires my superior sorting skills! 

No. I am a total amateur - never bought tons of bricks - just have my piles ... and then I simply do not care about subtle differences in parts - I am just using them.

So when Piqabrick can achieve that for the above average part collectors/sellers/re-sellers/people who care about that: Go for it!

I sure hope that this will be a success. Particularly when there is a need. Since it makes sense. That is the most important thing to me.

Cross my fingers and all the very best!

Thorsten

 

[MAB]

On 11/13/2019 at 3:43 PM, Calabar said:

Sometimes I needed to identify some parts from a mixed lot (usually in order to detect from some specific bricks the sets from which the parts came from and try to rebuild them) and this operation was very time consuming.

It shouldn't be too time consuming, if you know how to use bricklink searches efficiently.  You don't really even need to find rarer parts or prints to do this, common parts work if you have some that are stuck together.

For example, if you find two or three parts connected together, then the multiple search will help with this:

https://www.bricklink.com/catalogIn.asp?viewMulti=Y

Choose two or three parts that you can identify easily and enter both the part and colour for each, and you will get a list of the sets that those three parts appear in. Even for three relatively common bricks, the lists are surprisingly short. However, if you can identify one less common part, often the list is very short.

For example, if you find two of these

on top of one of these

then they appear to be relatively common parts. The lime green 1x1 slopes are in 82 sets and the white 2x2 slope is in 366 sets. 82 is a large number of sets to start looking through.

Yet together, those two quite common and easy to identify parts occur in only 8 sets and you can quickly and easily check the set pictures to spot that combination, or start to look for other common parts that they are connected to that will restrict the search. If you have five or six reasonably common parts connected together (and it is a real set) then chances are you can identify the set unambiguously using a couple of multiple parts searches.

Of course, this will only help if there are remnants of sets with some parts still connected.

 

[Calabar]

@Toastie

Eheh that's a very good point! If you don't need to exactly identify your bricks, probably this tool is not very useful!

@MAB

Thanks, this is quite useful and I didn't know it. I usually did separate searches in that case.
The cases I referred before anyway are often related to printed parts and identify a print is often difficult and requires time. Besides the existence of very similar parts with completely different IDs don't help.

[SylvainLS]

56 minutes ago, Calabar said:

Besides the existence of very similar parts with completely different IDs don't help.

Very similiar parts with very similar (but still different) IDs don’t help either  

[MAB]

3 hours ago, Calabar said:

Thanks, this is quite useful and I didn't know it. I usually did separate searches in that case.

The cases I referred before anyway are often related to printed parts and identify a print is often difficult and requires time. Besides the existence of very similar parts with completely different IDs don't help.

This is where experience / knowledge helps. Although the language used in part titles sometimes varies, knowing what search terms are likely to appear can make it easy to identify prints relatively quickly.

If I get anything like that and I don't know what it is, I usually identify the parent part type, then first go to colour, then part type, then see how many parts there are of that part type. If I can identify the shape of the part, I use that. If lots of parts are still possible, then I search for keywords, if not (so a page or less) then I go by eye. I can usually find most unidentified printed parts in under a minute.

[leafan]

I have high hopes for this project and hope it becomes widely used.

In the meantime, here is another Lego sorting project which uses AI:

I first heard about it in this article at The Register.