Giter Club home page Giter Club logo

Comments (5)

dhruvbatra avatar dhruvbatra commented on September 12, 2024 1

You're welcome! We're happy to hear of your (continued) interest.

Participants are free to use lower resolutions for their models/methods. We are simply specifying what resolution is available at test time.

from habitat-challenge.

edbeeching avatar edbeeching commented on September 12, 2024

Thank you for such a prompt response. My concern is if I, for example, train with the simulator running at 128x96 and then test at 640x480 and manually resize with OpenCV, the anti-aliasing/downsampling may not achieve the same result. Perhaps my concern is unfounded and the model will be robust to this change. Time will tell.

I think last year other config files were available at test time? or perhaps I am getting mixed up with another competition.

from habitat-challenge.

dhruvbatra avatar dhruvbatra commented on September 12, 2024

No, that's a reasonable question.

train with the simulator running at 128x96 and then test at 640x480 and manually resize with OpenCV, the anti-aliasing/downsampling may not achieve the same result.

Yes, that could lead to differences. I would expect them to be minor. I assume you don't want to train with 640x480 --> downsized to 128x96 because that would be slower to train?

I think last year other config files were available at test time?

I don't think so, but perhaps @mathfac or @abhiskk can confirm.

from habitat-challenge.

mathfac avatar mathfac commented on September 12, 2024

@edbeeching, previous year challenge was run with a fixed resolution 256x256, you can check it here.

Training with 640x480 instead of 128x96 shouldn't hit performance significantly, as a render call will be needed anyway.
I would advice next options:

  1. Train with 640x480 and downsample to 128x96 during training to have same frame transformation.
    Or
  2. Train with 128x96, and run evaluation on val split with 640x480 and downsampling and check that success metrics doesn't change much.

from habitat-challenge.

edbeeching avatar edbeeching commented on September 12, 2024

@mathfac
Thank you for confirmation, I must of been thinking of another competition.
I was expecting the rendering to be slower with the higher resolutions, as per the table on the habitat-sim github page, https://github.com/facebookresearch/habitat-sim#performance.

Thank you for the advice, I will compare the two options and see what works best.

from habitat-challenge.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.