AIGC-to-NFT

Abstract//

AI painting is considered one of the future directions in the field of artistic creation. At the same time, the form of NFT digital collectibles has also impacted the way of buying and selling collectibles. In the rapidly changing AI era, we intend to use AI to create Monet style portraits of film characters to sell as NFT to explore and integrate the two emerging trends. This project connects image to image translation and text to image translation through GPT, and tries to tap the new market of NFT digital art creation and application.

The PatchGAN discriminator consists of a sequence of convolution layers, which can be built using the downsampling blocks defined earlier. To optimize these parameters, we need to define the loss functions, including the adversarial loss, identity loss, and cycle loss.!

Step 1: Gathering Training Images

Step 2: Pre-Processing Training Images (Using DLIB package to resize the portrait)

Step 3: Training the dataset by using CycleGAN model

Step 4: Using VQGAN to Refine the Details

Step 5: Using GPT-3.5 Turbo as a serach engine to find Image