rozgo

@rozgo

Simulation and AI Engineer — I train robots inside virtual worlds to become autonomous intelligent systems in the real world.

Vrijeme pridruživanja: listopad 2008.

Tweetovi

Blokirali ste korisnika/cu @rozgo

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

  1. Prikvačeni tweet
    9. sij

    Real-time monocular depth prediction using a and pipeline. I use to have this pipeline in TensorFlow, but moved away from it.

    Poništi
  2. 3. velj

    Got a Karbon 700 from , a camera and sensors from . Working on a proof of concept for my real-time computer vision streaming pipeline—before I go inflicting this tech stack on other teams. Developed with Rust, GStreamer and PyTorch.

    Poništi
  3. 23. sij

    We can agree to disagree.

    Poništi
  4. 16. sij

    Ordered an Karbon700 Looking forward to some fun experiments.

    Prikaži ovu nit
    Poništi
  5. 16. sij

    Added semantic segmentation to the real-time computer vision streaming pipeline. Now in real need of a native Tensor media type in GStreamer if I want to stay above 100FPS.

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet

    A 4 year old child actually has a few hundred million years of experience, not 4. Their rapid learning/generalization is much less shocking/magical considering this fact.

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    12. pro 2019.

    All of us writing C and C++ are living on borrowed time. The only safe future is Rust. Prepare your code to go out of scope.

    Prikaži ovu nit
    Poništi
  8. proslijedio/la je Tweet
    23. lis 2019.

    Se deben formar y crear laboratorios de investigación que realmente lo sean, y cambiar la currícula educativa. A veces se quiere replicar modelos, pero son otras realidades. Si no cambia, seguiremos en el mismo cículo, Tendencias

    Poništi
  9. 11. ruj 2019.
    Prikaži ovu nit
    Poništi
  10. 11. ruj 2019.

    AI policy is defined as public policies that maximize the benefits of AI, while minimizing its potential costs and risks.

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet
    26. kol 2019.

    We are releasing our work on goal-directed visual navigation. We introduced a method that harnesses different perception skills based on situational awareness. It makes a robot reach its goals more robustly and efficiently in new environments.

    Poništi
  12. 3. kol 2019.

    Amazing UE4 developer, audio genius and one of the most talented people I’ve worked with.

    Poništi
  13. 10. srp 2019.

    We’ll need to teach it, cause , on its own, won’t develop empathy for the painful privilege of our existence.

    Poništi
  14. 8. srp 2019.

    Very early results of decoding synthetic characters into real humans using , and GANs model. Part of our quest to bridge the reality gap in real time synthetic data pipelines.

    Poništi
  15. proslijedio/la je Tweet
    11. lip 2019.

    Weight Agnostic Neural Networks 🦎 Inspired by precocial species in biology, we set out to search for neural net architectures that can already (sort of) perform various tasks even when they use random weight values. Article: PDF:

    Prikaži ovu nit
    Poništi
  16. 27. svi 2019.

    Encoding mechanics and behaviors into economy primitives, effectively creating an economy algebra of the problem domain, has the added benefit of improving sample efficiency of reinforcement learning models by providing a fully dense reward state space.

    Poništi
  17. 24. svi 2019.

    Testing turning a semantically labeled scene into photo-realistic video. Lacking a bit of the "photo-real" in this first test, but cool nonetheless.

    Poništi
  18. 23. svi 2019.

    Multi-robot coordination: Distributed synchronization of industrial robots through ROS 2

    Poništi
  19. proslijedio/la je Tweet
    14. ožu 2019.

    A map from the unit square to the unit circle, without trigonometry, that I found in this paper . I reduced and simplified to just v = maxcomp(abs(v))*normalize(v) Not for sampling, but great for geometry. Presumably works in 3D!

    Prikaži ovu nit
    Poništi
  20. 5. ožu 2019.

    One way to accelerate a massive multiagent training simulation session is to get a human-in-the-loop to reinforce. Human actions are worth more than a thousand tries.

    Poništi

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:

    ·