selim

Cursed StyleGAN2 people after a deep style transfer, makes some interesting thingspic.twitter.com/DYmJ7zPuDJ

"One of the big mysteries of #sleep is why it is necessary to go offline and be unconscious." https://science.sciencemag.org/content/367/6476/366 … @sciencemagazine by William Wisen and Nicholas Franks @Imperialcollege @ImperialUKDRI w/ @ScienceVisuals @oooitsplasmapic.twitter.com/DJEypbSV27

Music sketch diary #2. Prototyping. Generate music and synthesizers by sketching wavetables and scores. #openFrameworks #sonification #creativecoding #generative #synthesizerpic.twitter.com/UI9xMOQAbu

#popcornpic.twitter.com/1c8u9yAvyN

Long exposure of a plane taking offpic.twitter.com/giLGYMW14s

This animation is so pleasing to watchpic.twitter.com/6FJNqe5mKO

Visualization of a net learning a 2D image function with four different random start parameters. Interesting how much the character of the init remains after 1200 updates.pic.twitter.com/lRwFGD9y4Z

Entered @facebookai’s deepfake detection challenge from first place Good luck to all! Going to sleep now https://twitter.com/cristiancanton/status/1204809857762766848 …pic.twitter.com/5aPHLvAlqR

Powerful opening #NeurIPS2019 keynote by @celestekidd! Many inspirational thoughts from developmental psychology. Curiosity and intrinsic motivation in RL have a lot of work to do.pic.twitter.com/HpUZbCkNwl

How do you make a VAE learn informative latent representations? The normal prior over-regularises the posterior distribution. Our idea: learn the prior, thus creating more expressive, hierarchical latent spaces. See our blog https://argmax.ai/blog/vhp-vae/  #NeurIPS2019pic.twitter.com/31syYdaHL0

Unsupervised pre-training now outperforms supervised learning on ImageNet for any data regime (see figure) and also for transfer learning to Pascal VOC object detection https://arxiv.org/abs/1905.09272v2 …pic.twitter.com/cciL5Db73x

#pico8 cls() ::_:: for i=0,1600 do if(i<15)pal(i,({0,128,130,2,136,8,142,137,9,10,135,7})[i+1],1) x=rnd(128) y=rnd(128) a=atan2(x-64,y-64)+.17 d=rnd(7) pset(x+cos(a)*d,y+sin(a)*d/3-cos(a)*d/4,max(0,pget(x,y)+.87-rnd())) end circfill(64,64,5,11) flip()goto _pic.twitter.com/0kQiEvFvii

Note how the tail wagging ceases, so as to divert more neural activity to their processor cores, as their reality shatters. https://bit.ly/2DU70c3 pic.twitter.com/Sp73rB6Siy

Glad to share our new work, #StarGANv2. Our model can generate high-quality images reflecting the diverse styles (e.g., hairstyles, makeup) of reference images. arXiv: https://arxiv.org/abs/1912.01865  github: https://github.com/clovaai/stargan-v2 … co-authors: @YoungjungUh @Jaejun_Yoo @JungWooHa2pic.twitter.com/AHJx4itYLL

pic.twitter.com/0TQGM1JwmR