-Fine-tuned StyleGAN on custom data set for AI Image Generation -Developed an automated pipeline for data pre-processing and cleaning, incorporating techniques like extraction and Super Resolution etc.-Utilizing a pre-trained StyleGAN model, conducted inference and classification with a multi-ethnicity classifier to analyze and categorize large-scale data into distinct ethnicity clusters. This process also involved evaluating the centroid for latent inversion of these clusters and optimizing the centroid latent vector of a particular ethnicity with respect to other ethnicity latent vector to mimic the skin tone of the facial identity of the centroid latent vector-Implemented the overall flow of training the PsP encoder to achieve the latent inversion of real identity.-Conducted research to develop an innovative unsupervised technique for disentangling the facial identity of generated GAN images-Successfully trained the Face Identity Disentanglement Model to transfer the attributes of a real identity face onto an AI-generated image-Employed state-of-the-art Generative Diffusion models to generate naked busts from garment busts through inpainting, effectively expanding the dataset for our project.-Leveraged alpha pose to accurately determine and classify diverse human poses.