Spatial AI Ecologies explores creative AI in a collaborative fashion to produce a new cognition of interiority. Rooted in the material basis of intelligence, and how geology, biology and technology interact, the work utilizes ai algorithms to construct a new intriority; an amalgam of media created using AI.
The following algorithms have been used:
-
3D GAN: Following this paper Improved Adversarial Systems for 3D Object Generation and Reconstruction we built a GAN model that generates AI 3D meshes. Check michleangelo.
-
AI Generated music: We used Open AI Jukebox to generate music. The notebook can be found here: jukebox.ipynb and here you can find samples of the results.
-
Text to image: We used Open AI DALL-E to generate images from text. The notebook can be found here: text_to_image_dalle_ipynb.ipynb.
-
MeshRCNN: MeshRCNN can predict 3D objects from 2D images. The notebook can be found here: MeshRCNN.ipynb.
-
PIFuHD: We used PIFuHD: Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization to reconstruct 3D representation object from of a 2D image. The notebook can be found here: PIFuHD.ipynb.
-
VQGAN+CLIP: We used VQGAN and CLIP to generate images from a text prompt. The notebook can be found here: VQGAN+CLIP.ipynb. Results can be found here.
https://drive.google.com/drive/folders/1o1fAlJd13Uiad-jfx7taFNEeRNG_emK0?usp=sharing

