Can you GAN?
summary
Shows working GAN code using Apache MXNet bindings for Clojure, with a generator and discriminator trained on roughly 1,000 scraped flan images (starting at 28×28, up to 128×128) producing convincing and bizarre outputs; training uses deconvolution/convolution layers, leaky ReLU, image record iterators and small batch sizes (~4–5 on 128px). MXNet Clojure support is merged and jars are on Maven. Notes practical limits (unstable outputs, stopped at epoch 161) and points to Google BigGAN 256×256 Colab demos.
AI-written summary · about
chapters (40)
- 0:00Intro
- 0:18flan
- 2:59Cat Pie
- 3:16Cat Show of Hands
- 4:11GAN Images
- 5:51What is GAN
- 6:06The Network
- 7:12Activation Functions
- 7:48Why GAN
- 8:07universality
- 8:28Legos
- 8:48Learning
- 9:41Deep Learning
- 11:07Rayleigh
- 11:19The GAN
- 12:53The Generator
- 13:55Can we gain
- 16:38Lets gain
- 16:50Ingredients
- 17:11flan pictures
- 17:28scraper
- 18:03prepare
- 20:32discriminator
- 21:45generator
- 22:24image iterator
- 23:18cooking show
- 23:33epic zero
- 23:49Abuk
- 24:11Its Learning
- 24:26Its Surreal
- 25:03Its On Fire
- 25:28Green Flan
- 26:56Auction
- 27:35Price
- 28:00Big Gann
- 28:34Cheeseburger
- 29:34Dog and Frog
- 30:10Wrap up
- 30:31Ganz can help
- 30:54Closure
related talks
37mHarnessing the power of Generative Adversarial Networks (GANS) for supervised learning
40mVisual concept learning from few images
29mDeep Learning Needs Clojure
44mMachine Learning with Artificial Neural Networks in Clojure
40mOverview of Generative Adversarial Networks (GANs)
01h 00mTutorial on Deep Learning IV
43mHitchhiker’s Guide to Generative Adversarial Networks (GANs)
01h 27mMaking Faces: Conditional Generation of Faces using GANs...