Comics
Fourth NFT Collection
I base my art collections on personal interests and intrigue of how an Artificial Intelligence will generate the images. My past collections have been fairly easy for the Machine Learning algorithm to train and generate results, since they were fairly similar objects like flowers or abstract, where the results would be interesting no matter what. So I decided to make kick things up a notch and use a dataset where the images would be vastly different.
​
I realized that this collection would pose some challenges, like having enough images in the dataset to get clear results and longer training times then my previous ones. I go into more detail into issues and challenges posed in each generation.
Gen 1
Image Dataset #: 6992
Image Size: 1024 x 1024
Seeds : 1 - 1000
Total NFTs: 5
Minted: 5 / 5
Training Time: 11 Days
The First Generation I was actually debating on not even turning into NFTs since it ended up collapsing and overfitting the images. Which means a lot of the images looked similar and eventually the results turned into blobs. There were some interesting results so I made a small collection of 5 NFTs.
Gen 2
Image Dataset #: 13000
Image Size: 1024 x 1024
Seeds : 1 - 1000
Total NFTs: 15
Minted: 15 / 15
Training Time: 20 Days
The Second Comic Generation created with a larger image dataset of 1300 images and training for 20 days. I started this training from scratch since my first model from Gen 1 collapsed and I changed a few parameters.
Gen 3
Coming In The Future