Original source: Jason, founder of BuilderDao
< b>BlockBeats Note: The NFT copyright detection tool Yakoa is mainly used to index NFT items, and uses artificial intelligence technology to analyze the similarity of pictures to determine whether the picture is pirated. In fact, this technology has been widely used in web2. Jason Chen, the founder of BuidlerDAO, conducted a simple science popularization on Twitter to learn about the application scenarios of this technology in Web3 and how it works. BlockBeats organizes the content as follows:
Talk about the NFT piracy detection tool that received a $4.8 million seed round yesterday Yakoa, Currently the Chinese media call it "fraud detection", but I think this term is not accurate< /b>, it is estimated to be a direct machine translation. Its core problem is to judge whether there are pirated and infringing NFT projects through image similarity detection. It has won the entry of many star institutions, including Uinswap, which should also be because of its acquisition of genie s reason.
Picture similarity detection is already a very mature technology in web2. The common ones are Baidu's image search, or Taobao's photo identification. It is quite a tricky way to bring this set of technologies into web3. We can see through the official website that it lists two types of cases, namely brand (Nike and Marvel) and project (Gucci bear and Azuki) , The scene of brand customers still exists.
For example, Marvel can use the Yakoa tool to see how many NFTs have used logos without authorization, which solves the problem of "visibility", but still cannot solve the problem of "management". After all, even if the other party's infringement already exists It is also impossible to interfere with the other party’s destruction of things on the chain. At most, it can only be used as evidence to require the trading platform to block it. Even if someone can be found through traditional methods to send him a lawyer’s letter, so what?
The same is true for the second type of project. As shown in the figure above, all similar Azuki with hat characteristics can be retrieved. I am ashamed to say that I really have one of the bzuki below, so I can conclude that other Behind it is the process of using a specific eigenvalue to search for an image. It is technically not difficult. I think the more difficult thing is the eigenvalue such as the Marvel logo, which needs to traverse all the items generated on the entire network.
Moreover, it still supports multiple chains. At present, tens of thousands of projects have been generated, and each project has 10,000 pictures. This traversal process consumes a lot of server resources.
Although it is a mature web2 technology, we can also see some original innovations of web3 from its development documents. As shown in the figure below, it first defines assets. In his eyes, assets and tokens are decoupled. We usually The NFT mentioned actually integrates assets and tokens, that is, if I hold Azuki#9527, its essence is a token with an id of 9527, but we will bring the picture attached to it by default.
But the scene of Kayoa is to search for pictures, so in his eyes, the picture attached to the token is more like an asset. In fact, this is the essence of NFT. The "meaningless" number 12345 is linked to the chain As shown in the picture below, it deconstructs the asset model into the chain, address, tokenid and original block, that is, the block where the asset first appeared, so that we can judge its birth sequence on the chain time.
The following figure is a case of running. Call the interface to set the specific NFT to be detected. After entering the chain, contract address, and tokenid, N results will be returned according to the format of the asset model, so you can set It is understood as searching for images by image in the NFT field. I think the specific scenario has not been explored yet. After all, as mentioned above, it can be seen but how to manage it? After investing in Uinswap, I don’t know how to apply it to genie, so wait and see.
< br>
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia