Text to FigureText → Imageposter

NC-TTT: Noise-Contrastive Test-Time Training Poster

A conference poster presenting NC-TTT, a method using noise contrastive estimation for test-time training. It details the Y-shaped architecture, motivation regarding kernel density estimation, and experimental results on CIFAR-10-C and VisDA-C datasets showing significant accuracy improvements.

Paper context

Paper title: NC-TTT: A Noise Contrastive Approach for Test-Time Training Abstract: A conference poster presenting NC-TTT, a method using noise contrastive estimation for test-time training. It details the Y-shaped architecture, motivation regarding kernel density estimation, and experimental results on CIFAR-10-C and VisDA-C datasets showing significant accuracy improvements. Paper body (method & results): <!DOCTYPE html> <html lang="en" prefix="dcterms: http://purl.org/dc/terms/"> <head> <meta content="text/html; charset=utf-8" http-equiv="content-type"/> <title>NC-TTT: A Noise Contrastive Approach for Test-Time Training</title>   <meta content="width=device-width, initial-scale=1, shrink-to-fit=no" name="viewport"/> <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/css/bootstrap.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv-fonts.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/latexml_styles.css" rel="stylesheet" type="text/css"/> <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/html2canvas/1.3.3/html2canvas.min.js"></script> <script src="/static/browse/0.3.4/js/addons_new.js"></script> <script src="/static/browse/0.3.4/js/feedbackOverlay.js"></script> <base href="/html/2404.08392v1/"/></head> <body> <nav class="ltx_page_navbar"> <nav class="ltx_TOC"> <ol class="ltx_toclist"> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S1" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">1 </span>Introduction</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S2" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2 </span>Related work</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S3" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3 </span>Methodology</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S3.SS1" title="In 3 Methodology ‣ NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.1 </span>The proposed method</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S3.SS2" title="In 3 Methodology ‣ NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2 </span>Noise-contrastive Test-time Training</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S3.SS3" title="In 3 Methodology ‣ NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3 </span>Selecting the distribution variances</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S4" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4 </span>Experimental Settings</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S5" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5 </span>Results</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S5.SS1" title="In 5 Results ‣ NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.1 </span>Image classification on common corruptions</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S5.SS2" title="In 5 Results ‣ NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.2 </span>Image classification on sim-to-real domain shift</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#S6" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6 </span>Conclusions</span></a></li> <li class="ltx_tocentry ltx_tocentry_appendix"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#A1" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">A </span>Deriving the posterior of Equation (6)</span></a></li> <li class="ltx_tocentry ltx_tocentry_appendix"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#A2" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">B </span>Results on different levels of CIFAR-10-C corruptions</span></a></li> <li class="ltx_tocentry ltx_tocentry_appendix"><a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#A3" title="In NC-TTT: A Noise Contrastive Approach for Test-Time Training"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">C </span>Hyperparameter search on VisDA-C</span></a></li> </ol></nav> </nav> <div class="ltx_page_main"> <div class="ltx_page_content"> <article class="ltx_document ltx_authors_1line"> <h1 class="ltx_title ltx_title_document">NC-TTT: A Noise Contrastive Approach for Test-Time Training</h1> <div class="ltx_authors"> <span class="ltx_creator ltx_role_author"> <span class="ltx_personname">David Osowiechi &Gustavo A. Vargas Hakim<span class="ltx_note ltx_role_footnotemark" id="footnotex1"><sup class="ltx_note_mark">1</sup><span class="ltx_note_outer"><span class="ltx_note_content"><sup class="ltx_note_mark">1</sup><span class="ltx_note_type">footnotemark: </span><span class="ltx_tag ltx_tag_note">1</span></span></span></span> &Mehrdad Noori &Milad Cheraghalikhani &Ismail Ben Ayed &Christian Desrosiers </span><span class="ltx_author_notes">Equal contribution</span></span> </div> <div class="ltx_abstract"> <h6 class="ltx_title ltx_title_abstract">Abstract</h6> <p class="ltx_p" id="id2.id1">Despite their exceptional performance in vision tasks, deep learning models often struggle when faced with domain shifts during testing. Test-Time Training (TTT) methods have recently gained popularity by their ability to enhance the robustness of models through the addition of an auxiliary objective that is jointly optimized with the main task. Being strictly unsupervised, this auxiliary objective is used at test time to adapt the model without any access to labels. In this work, we propose Noise-Contrastive Test-Time Training (NC-TTT), a novel unsupervised TTT technique based on the discrimination of noisy feature maps. By learning to classify noisy views of projected feature maps, and then adapting the model accordingly on new domains, classification performance can be recovered by an important margin. Experiments on several popular test-time adaptation baselines demonstrate the advantages of our method compared to recent approaches for this task. The code can be found at: <a class="ltx_ref ltx_url ltx_font_typewriter" href="https://github.com/GustavoVargasHakim/NCTTT.git" title="">https://github.com/GustavoVargasHakim/NCTTT.git</a></p> </div> <section class="ltx_section" id="S1"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">1 </span>Introduction</h2> <div class="ltx_para ltx_noindent" id="S1.p1"> <p class="ltx_p" id="S1.p1.1">A crucial requirement for the success of traditional deep learning methods is that training and testing data should be sampled from the same distribution. As widely shown in the literature <cite class="ltx_cite ltx_citemacro_cite">Recht et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib1" title="">2018</a>); Peng et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib2" title="">2018</a>)</cite>, this assumption rarely holds in practice, and a model’s performance can drop dramatically in the presence of domain shifts. The field of Domain Adaptation (DA) has emerged to address this important issue, proposing various mechanisms that adapt learning algorithms to new domains.</p> </div> <div class="ltx_para ltx_noindent" id="S1.p2"> <p class="ltx_p" id="S1.p2.1">In the realm of domain adaptation, two notable directions of research have surfaced: Domain Generalization and Test-Time Adaptation. Domain Generalization (DG) approaches <cite class="ltx_cite ltx_citemacro_cite">Volpi et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib3" title="">2018</a>); Prakash et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib4" title="">2019</a>); Zhou et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib5" title="">2020</a>); Kim et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib6" title="">2022</a>); Wang et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib7" title="">2022</a>)</cite> typically train a model with an extensive source dataset encompassing diverse domains and augmentations, so that it can achieve a good performance on test examples from unseen domains, without retraining.</p> </div> <div class="ltx_para ltx_noindent" id="S1.p3"> <p class="ltx_p" id="S1.p3.1">Conversely, Test-Time Adaptation (TTA) <cite class="ltx_cite ltx_citemacro_cite">Wang et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib8" title="">2021</a>); Khurana et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib9" title="">2021</a>); Boudiaf et al. (<a class="ltx_ref" href="https://arxiv.org/html/2404.08392v1#bib.bib10" title="">2022</a>)</cite> entails the dynamic adjustment of the model to test data in real-time, typically adapting to subsets of the new domain, such as mini-batches. TTA presents a challenging, yet practical problem as it functions without supervision for test samples or access to the source domain data. While they do not require training data from diverse domains as DG approaches, TTA methods are often susceptible to the choice of unsupervised loss used at test time, a factor that can substantially influence

The prompt

Above I've shared:
(1) the full paper text,
(2) all paper figures labeled by figure number,
(3) the caption for the central poster figure I'm building.

TASK: This is a CONFERENCE POSTER. **NOT** an academic-paper figure.
Style requirements:

  - Multi-section layout with a clear poster structure: large title banner
    at the top with the paper title + author/affiliation strip, then 3-6
    distinct content panels arranged in columns or a grid.
  - Large legible fonts (text must be readable at 2 m viewing distance) —
    headings ≥ 60 pt visual size in the final image.
  - Use colour blocks / panel backgrounds to delineate sections (this is
    what makes it a poster, not a single-figure diagram).
  - Aspect ratio: portrait or landscape rectangle, NOT square.

If your output looks like a standard academic-paper figure (single panel,
no title banner, dense small text, no colour blocks), you've failed the
task. Render the COMPLETE poster, not just the central figure.

Just give me the final poster image.

Try this prompt now

Open it inside the generator with the prompt pre-filled.

Try this prompt

Related prompts

Text to Figure

NC-TTT: Noise-Contrastive Test-Time Training Poster

Paper context

The prompt

Try this prompt now

Related prompts

PINN Framework Pipeline for Microscopy Data

DP-FTRL Privacy Amplification Analysis — Conference Poster

Pref-CTRL: Value Function Training Overview