Dance of the Ancestors...
summary
Curates 11,000 curated mask images and trains a GAN that generates novel African-mask images, then upsamples 64× outputs with a GAN + super-resolution GAN pipeline to about 378 px (6×) producing texture and detail not present in bicubic upsampling. Runs tens to hundreds of TPU-accelerated experiments and reports on image quality, novelty, and data-impact findings, plus a web inspection tool implementing semantic nearest-neighbor search to compare generated images to the dataset.
AI-written summary · about
chapters (5)
- 0:00Start
- 1:39The Masquerade Dances
- 6:53Generative Adversarial Networks
- 14:01Results
- 15:27Algorithmic Art Inspection
related talks
12mENHANCE! Upscaling Images with Neural Networks
37mHarnessing the power of Generative Adversarial Networks (GANS) for supervised learning
01h 27mMaking Faces: Conditional Generation of Faces using GANs...
03mBuilding artificial people: Endless possibilities and the dark side
28mBackground removal without background knowledge
15mDrag Makeovers, via AI!
47mUnmasking Generative Adversarial Networks
32mCan you GAN?