Carl Doersch

@CarlDoersch

Researcher at DeepMind

Vrijeme pridruživanja: travanj 2017.
1 fotografija ili videozapis Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @CarlDoersch

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @CarlDoersch

  1. 9. pro 2019.

    The self-supervised dream is slowly coming true

    Very happy to share our latest unsupervised representation learning work! In addition to SOTA linear classification, we beat supervised networks on ImageNet with 2-5x less labels and transfer to PASCAL detection better than supervised pre-training.
    Prikaži ovu nit
    1 proslijeđeni tweet 9 korisnika označava da im se sviđa
    Poništi
  2. 5. srp 2019.

    Training on only synthetic data, our new paper gets near the SOTA on the real-world 3DPW human pose dataset by focusing on motion. Simulation is underappreciated for representation learning; you just have to keep the nets from overfitting to the domain.

    70 proslijeđenih tweetova 304 korisnika označavaju da im se sviđa
    Poništi
  3. proslijedio/la je Tweet
    23. svi 2019.

    Getting closer to the dream! A network that uses unlabelled images to boost performance when labels are scarce (new SOTA), and it's no worse than ResNet when labels are plentiful. Also: Unsupervised net + just a linear on top outperforms original AlexNet!

    5 replies 115 proslijeđenih tweetova 405 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  4. proslijedio/la je Tweet
    8. tra 2019.

    Excited to share our newest work (& my first final-author paper!) led by Victor Bapst and Alvaro Sanchez-Gonzalez (w/ ). We investigate structured, object-centric agents that can build things!👷‍♀️🏗

    6 replies 40 proslijeđenih tweetova 142 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    2. kol 2018.

    Presenting a novel hard attention mechanism for Visual QA (VQA-CP and CLEVR) which retains only the small number of visual features needed to answer the question. Reduced processing allows us to train larger relational models and achieve 98.8% on CLEVR:

    180 proslijeđenih tweetova 559 korisnika označava da im se sviđa
    Poništi

@CarlDoersch još nije objavio nijedan Tweet.

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·