1. What does "Learning to see like children" mean ?
    Nowadays in computer vision there exists a huge number of applications which perform very well in specific contexts. Yet, what is still missing is a system which can approach the problem of vision in general, continuously processing videos, receiving supervisions and producing predictions to describe the scene. We want to build such a general-purpose agent.

  2. Can I try DVA on my own videos ?
    Of course ! You are welcome to test DVA and report us your impressions, criticisms and share the results of your experiments. See the Download section for more details.

  3. I have seen you can give rates to your agents. How does it all work ?
    DVA is a general purpose system, which has not been conceived to work on specific video categories, nor for specific tasks and applications. In the Rating section of this website, after registering, you will be able to rate the behavior of several agents, for which you will observe some videos demonstrating their performance. This is a brand new approach to computer vision evaluation, which does not rely any more on collections of data sets, which too often create a tremendous bias in computer vision applications and limit the generality of the systems.