Stefan Scholz

📣 Happy to announce the publication of our article (w @nilsweidmann.bsky.social, @friederikeq.bsky.social, @sebnagel.bsky.social, @yannistheocharis.bsky.social & Molly Roberts) on the complexity and availability of community guidelines @icwsm.bsky.social! 🔗 ojs.aaai.org/index.php/IC...

🚨🚀 Looking for a comparative dataset on social media platforms? We’re excited to launch COMPARE! This is a collaborative effort by @nilsweidmann.bsky.social , @friederikeq.bsky.social , @sebnagel.bsky.social , @yannistheocharis.bsky.social & Molly Roberts. 🧵⤵️ (1/5)

Sharing our new preprint "An Image is Worth K Topics: A Visual Structural Topic Model with Image Embeddings" with @mansmag.bsky.social @matmagnani.bsky.social Alexandra Segerberg and Nataša Sladoje. Available on ArXiv: arxiv.org/abs/2504.10004

Postdoctoral Position in Computational Social Science Deadline: 15.03.2025

Want to work with us? stellen.uni-konstanz.de/jobposting/8...

Now in FirstView: “Improving Computer Vision Interpretability: Transparent Two-Level Classification for Complex Scenes.” @stefan-scholz.bsky.social, @nilsweidmann.bsky.social‬, @zacharyst.bsky.social, @keremoglu.bsky.social, and Bastian Goldlücke propose a two-level method for image classification.

Finally out in @polanalysis.bsky.social (w/ @stefan-scholz.bsky.social, @zacharyst.bsky.social, @keremoglu.bsky.social and Bastian Goldlücke): "Improving Computer Vision Interpretability: Transparent Two-Level Classification for Complex Scenes" Available #OpenAccess at doi.org/10.1017/pan....

Protest Segments - a Hugging Face Space by ciass Discover amazing ML apps made by the community

Do you want to try out the method with your own images? Here is our free demo application. huggingface.co/spaces/ciass...

The novelty of this method is that it provides new insights for comparative politics: While persons, flags and signboard are important objects in protest images, particular features of protest differ across countries and protest episodes. Our method can detect these.

Our method detects objects present in images, creates feature vectors from those objects and uses them as input for machine learning classifiers. We tested this on a new dataset of 140k images to predict which ones show protest. The accuracy is roughly on par with popular CNNs.

While existing classifiers for images reach high levels of accuracy, it is difficult to systematically assess the visual features on which they base their classification. This problem is especially pressing for complex images that contain many different types of objects.

Improving Computer Vision Interpretability: Transparent Two-level Classification for Complex Scenes Treating images as data has become increasingly popular in political science. While existing classifiers for images reach high levels of accuracy, it is difficult to systematically assess the visual f...

New paper forthcoming in PA, w/ @nilsweidmann.bsky.social, @zacharyst.bsky.social, @keremoglu.bsky.social and Bastian Goldlücke! We propose a method that makes image classification more transparent by identifying which objects on images are related to the outcome. Preprint: arxiv.org/abs/2407.03786