Predicting the Politics of an Image Using Webly Supervised Data

Authors: Christopher Thomas, Adriana Kovashka

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally show that our method outperforms numerous baselines on both a large held-out webly supervised test set, and the set of crowdsourced annotations.
Researcher Affiliation Academia Christopher Thomas Adriana Kovashka Department of Computer Science University of Pittsburgh Pittsburgh, PA 15213 {chris,kovashka}@cs.pitt.edu
Pseudocode No The paper describes its method verbally and with a diagram, but does not include structured pseudocode or algorithm blocks.
Open Source Code Yes Our dataset, code, and additional materials are available online for download here: http://www.cs.pitt.edu/ chris/politics
Open Datasets Yes We propose and make available1 a very large dataset of biased images with paired text, and a large amount of diverse crowdsourced annotations regarding political bias. Our dataset, code, and additional materials are available online for download here: http://www.cs.pitt.edu/ chris/politics
Dataset Splits No The paper mentions training on a 'train set' and evaluating on a 'held-out test set' (75,148 images), but it does not explicitly state a separate validation dataset split with specific numbers or percentages.
Hardware Specification No The paper does not specify the exact hardware components (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper mentions software like PyTorch, Doc2Vec (implicitly Gensim), Text Boxes++, and Symspell, but it does not provide specific version numbers for these software components.
Experiment Setup Yes We train all models using Adam [34], with learning rate of 1.0e-4 and minibatch size of 64 images. We use cross-entropy loss and apply class-weight balancing to correct for slight data imbalance between L/R. We use an image size of 224x224 and random horizontal flipping as data augmentation. We use Xavier initialization [21] for non-pretrained layers.