Seminar CS717: Seminar on Causality and Neurosymbolic Learning Computer Vision (HWS 2025)

The Computer Vision seminar covers recent topics in computer vision. In HWS2025, the Computer Vision seminar focuses on “Causality and Neurosymbolic Learning in Computer Vision.”  This seminar explores recent works on how vision systems can go beyond pattern recognition by learning to reason and explain. Causality helps models understand cause-and-effect and make decisions that generalize better. Neurosymbolic Learning combines deep learning with symbolic rules, enabling vision models to work with concepts, logic, and reasoning similar to how humans think. Together, these ideas aim to make computer vision systems more robust, interpretable, and intelligent.

• This seminar is organized by Prof. Dr.-Ing. Margret Keuper
• Prerequisites: solid background in Machine Learning
• Maximum number of participants is 12 students

In this seminar, you will

• Read, understand, and explore scientific literature
• Summarize a current research topic in a concise report (10 single-column pages + references)
• Give two presentations about your topic (3 minutes flash presentation, 15 minutes final presentation)
• Moderate a scientific discussion about the topic of one of your fellow students
• Review a (draft of a) report of a fellow student

Please register via Portal2 by September 1.

If you're a Master student,  accepted into the seminar, please email your list of preferred reference papers (see attached list for Master Students) by September 13 (at least four choices) to Tejaswini Medi at tejaswini.medi@uni-mannheim.de.

The actual topic assignment will take place shortly afterward, and we will notify you via email.

Our goal is to assign one of your preferred areas of work. Please note that preferences will be allocated on a first come, first served basis.

• Seminar Schedule is provided here : Seminar Schedule (PDF, 190 kB)

Download the Seminar Report Template  from here : Latex Template

What is the goal of the report for master students?

For master students, the goal of the report is to demonstrate their understanding of the assigned paper within the context of the broader literature (which may include seminar topics). The emphasis should be on critical synthesis and contextualization.

After reading their report, the reader should:

• understand the paper’s main contributions and their significance without consulting other sources.
• be able to situate the paper within the literature, e.g., identify previous works it builds upon, problems it addresses, or relations to other approaches.

They are expected to read and provide references beyond their assigned paper.

What structure should the report have?

Structuring the report to achieve the goals above is a central part of the assignment. A good starting point is the structure of their presentation (or chapter/paper), complemented by an introduction and a conclusion/outlook.

Each student will be assigned a  following specific chapter by us from the textbook. The presentation and the accompanying report should focus on the fundamental concepts of the overall topic, with particular emphasis on the content of the assigned chapter.

• §2, §3: Introduction into Causality & Binary SCMs
• §4, §5: Learning Binary SCMs & Connections to ML
• §6.1 – §6.5: Multivariate SCMs (Part 1)
• §6.6 – §6.11: Multivariate SCMs (Part 2)
• §7: Learning Multivariate SCMs
• §8: Connections to ML
   

Textbook Reference : J. Peters, D. Janzing, and B. Schölkopf. Elements of causal inference: foundations and learning algorithms. The MIT Press, 2017.

Each student works on a topic within the area of the seminar along with an accompanying reference paper. Your presentation and report should explore the topic with an emphasis on the reference paper, but not just the reference paper.

We provide reference papers below .

• ProMark_Proactive_Diffusion_Watermarking_for_Causal_Attribution (PDF)  
• CausalVAE: Disentangled Representation Learning via Neural Structural Causal Models (PDF)
• CAUSALADV: ADVERSARIAL ROBUSTNESS THROUGH THE LENS OF CAUSALITY
• CausalPC: Improving the Robustness of Point Cloud Classification by Causal Effect Identification (PDF)
• Causality Is Key to Understand and Balance Multiple Goals in Trustworthy ML and Foundation Models
• CausVSR: Causality Inspired Visual Sentiment Recognition (PDF)
• Vision-and-Language Navigation via Causal Learning 

• Textbook : J. Pearl. Causality. Cambridge University Press, 2 edition, 2009.
• Tutorial on Neuro Symbolic Reasoning and Learning at AAAI Conference 2023