Mario Klingemannಪರಿಶೀಲಿಸಿದ ಖಾತೆ

@quasimondo

Artist, Coder, Data Collector, Archivist, Speaker, Obsessive Compulsive Orderer, Artist in residence

Munich, Germany
ಏಪ್ರಿಲ್ 2007 ಸಮಯದಲ್ಲಿ ಸೇರಿದ್ದಾರೆ

@quasimondo ತಡೆಹಿಡಿಯಲಾಗಿದೆ

ನೀವು ಖಚಿತವಾಗಿಯೂ ಈ ಟ್ವೀಟ್‌ಗಳನ್ನು ನೋಡಲು ಬಯಸುವಿರಾ? ಟ್ವೀಟ್‌ಗಳನ್ನು ನೋಡುವುದು @quasimondo ಅವರನ್ನು ತಡೆತೆರವುಗೊಳಿಸುವುದಿಲ್ಲ.

  1. ಪಿನ್ ಮಾಡಿದ ಟ್ವೀಟ್
    ಸೆಪ್ಟೆಂ 21,2016

    Peter sees the computer. "But the machine only creates what humans have taught it to," says Peter. "So do you," says Mummy.

  2. 2 ಗಂಟೆಗಳ ಹಿಂದೆ

    Good models copy. Great models steal.

  3. 2 ಗಂಟೆಗಳ ಹಿಂದೆ

    I tried to take some precautions so I don't fool myself and just replicate the training data, but I must admit, this looks almost too good.

  4. 5 ಗಂಟೆಗಳ ಹಿಂದೆ

    If want to hear a few more details about making art with neural networks - on March 4th I'll speak at

  5. 10 ಗಂಟೆಗಳ ಹಿಂದೆ

    And like every morning the neurographer visits his models to check what they have learned in the night.

  6. 23 ಗಂಟೆಗಳ ಹಿಂದೆ

    Nose problems? I have not the slightest idea what you are talking about.

  7. ಫೆಬ್ರ 1

    Oh noes, I've created a monster. This model trained on photos has fallen into the very-bad-fan-art valley and can't get out.

  8. ಫೆಬ್ರ 1

    Whilst I miss any mention of my 2006 Image Foam technique in the "previous methods" to solve the same problem, this is brilliant.

  9. ಫೆಬ್ರ 1

    The winners of the trailer park's "If I were President" fan-art competition have been announced:

  10. ಜನ 31

    As we say in the trade: garbage in, garbage out. Like before the face on the right is generated from the sketch on the left.

  11. ಜನ 31

    It has not yet learned how to shave properly. Maybe that also explains the ears.

  12. ಜನ 31

    Does it generalize or just memorize every image is has been trained on? Well, these cyberclopses are definitely not in the training set.

  13. ಜನ 30

    Generating faces from a sketch. I trained a pix2pix net on 1500 faces. Left input, right output.

  14. ಜನ 30

    A quick test for a new model. pix2pix is pretty close to magic. This is a snapshot after about 20 minutes of training on 1500 faces:

  15. ಜನ 29

    I can make movies with it, too, but the latent space is very erratic so the slightest changes in pose can cause a big leap. Anyway:

  16. ಜನ 29

    Some of those sets it creates are out of this world.

  17. ಜನ 29

    Training in progress...

  18. ಜನ 29

    Some experimentation with limbs and symmetry.

  19. ಜನ 28

    Neurography - the process of framing and capturing images in latent spaces. The Neurographer controls locations, subjects and parameters.

  20. ಜನ 28

    It has begun to learn faces. And shoes. (#4 might be on the edge of NSFW, but I can't spot any unsuitable parts)

  21. ಜನ 28

    Since my model keeps on training and I deliberately don't keep a backup of the current state I can never create the same image again.

ಲೋಡಿಂಗ್ ಸಮಯ ಸ್ವಲ್ಪ ತೆಗೆದುಕೊಳ್ಳುತ್ತಿರುವಂತೆನಿಸುತ್ತದೆ.

Twitter ಸಾಮರ್ಥ್ಯ ಮೀರಿರಬಹುದು ಅಥವಾ ಕ್ಷಣಿಕವಾದ ತೊಂದರೆಯನ್ನು ಅನುಭವಿಸುತ್ತಿರಬಹುದು. ಮತ್ತೆ ಪ್ರಯತ್ನಿಸಿ ಅಥವಾ ಹೆಚ್ಚಿನ ಮಾಹಿತಿಗೆ Twitter ಸ್ಥಿತಿಗೆ ಭೇಟಿ ನೀಡಿ.

    ಇದನ್ನೂ ಸಹ ನೀವು ಇಷ್ಟಪಡಬಹುದು

    ·