Neural Frens X Effect Network Partnership
We’re happy to announce that Neural Frens project is now an official partner of Effect Network! From now on we’re going to use Effect Force for labeling future Neural Frens collections that require classification.
What is Effect?
Effect is the first Blockchain-based framework for the Future-of-Work. In other terms, it’s a platform which lets you run various data structuring campaigns (i.e image labeling, translation and etc.) using Effect Force, a team of tested workers who willing to accomplish simple tasks in exchange for EFX (Effect native token) rewards.
Why do we need Effect Force?
Nowadays, most of NFT enthusiasts are quite used to generative NFT collections where every image has a set of traits with different rarity which also let you calculate the overall rarity of the given NFT. Such rarity is quite easy to calculate as every trait is usually drawn as a separate layer and then these layers are just shuffled using a special script. So, after the generation is over, it’s quite easy to calculate how many times each trait (layer) has been used.
However, it’s absolutely different for AI-generated collections. Each image GANs (generative adversarial networks) models produce is absolutely unique. You can’t just ask a StyleGAN to produce Pepe with hair or Pepe wearing a hat, unless you train a separate model to produce such results. Therefore, if you want to make an AI-generated collection with a rarity system similar to regular PFP NFT collections, you’re bound to do the generation first and only then determine repetitive patterns (potential traits) and manually process the dataset to assign these traits to certain images.
That’s how we did it with the Neural Pepe collection. First, generated 7,777 images and only then determined such traits as “Eyes”, “Background Color” and “Skin Color” to do the classification. Of course, manual classification of such a big amount of images entails low accuracy and a high chance misclassification (also, if you single-handedly sort through 7k neural frogs overnight they’re going to chase you in nightmares).
For that reason, some Neural Pepe ended up with misprints which we had to correct after a governance vote. Effect Force became an ideal solution for the Misprint Correction program. Instead of reclassifying NPs in a centralized way we were able to use Effect Force as an unbiased oracle to do the labeling.
The process was simple:
- We created a list of classification guidelines (pre-approved by the community)
- The guidelines were given to several dozens Effect Force workers
- Every image was labeled by 5 random Force workers who in our case were selecting the image’s background color
- Every image which received 3 or more votes for the same color was assigned with the color
Humanized Pepe
Recently, we released a collection of Humanized Pepe (H-NP) who are humanized twins of Neural Pepe. H-NP have their own rarity system and, again, they were labeled by Effect Force. Their traits will be released on the 16th of February.
Future Collaborations
Currently, Effect Team is planning to developing a special dApp which will allow us to run such labeling campaigns on a regular basis. Undoubtedly, it’s going to create a lot of cool opportunities for the future Neural Frens collections. For instance, we’re thinking of making a collection without pre-determined traits where it would be up to the community to discover traits and get them approved though the governance process and then assigned using Effect Force.
