| Title: | Eexplaining 3D semantic segmentation through generative AI-based counterfactuals |
|---|
| Authors: | ID Rozajac, Dzemail (Author)
ID Lukač, Niko (Author)
ID Schweng, Stefan (Author)
ID Gollob, Christoph (Author)
ID Nothdurft, Arne (Author)
ID Stampfer, Karl (Author)
ID Del Ser, Javier (Author)
ID Holzinger, Andreas (Author) |
|---|
| Files: |  Expert_Systems_-_2025_-_Rozajac_-_Explaining_3D_Semantic_Segmentation_Through_Generative_AI‐Based_Counterfactuals.pdf (27,14 MB)
MD5: 632F7495066012875B828A0201F916EE
|
|---|
| Language: | English |
|---|
| Work type: | Article |
|---|
| Typology: | 1.01 - Original Scientific Article |
|---|
| Organization: | FERI - Faculty of Electrical Engineering and Computer Science
|
|---|
| Abstract: | Interpreting the predictions of deep learning models on 3D point cloud data is an important challenge for safety-critical domains such as autonomous driving, robotics and geospatial analysis. Existing counterfactual explainability methods often struggle with the sparsity and unordered nature of 3D point clouds. To address this, we introduce a generative framework for counterfactual explanations in 3D semantic segmentation models. Our approach leverages autoencoder-based latent representations, combined with UMAP embeddings and Delaunay triangulation, to construct a graph that enables geodesic path search between semantic classes. Candidate counterfactuals are generated by interpolating latent vectors along these paths and decoding into plausible point clouds, while semantic plausibility is guided by the predictions of a 3D semantic segmentation model. We evaluate the framework on ShapeNet objects, demonstrating that semantically related classes yield realistic counterfactuals with minimal geometric change, whereas unrelated classes expose sharp decision boundaries and reduced plausibility. Quantitative results confirm that the method balances defined interpretability metrics, producing counterfactuals that are both interpretable and geometrically consistent. Overall, our work demonstrates that generative counterfactuals in latent space provide a promising alternative to input-level perturbations. |
|---|
| Keywords: | 3D point cloud, explainable artificial intelligence, counterfactual analysis, generative AI |
|---|
| Publication status: | Published |
|---|
| Publication version: | Version of Record |
|---|
| Submitted for review: | 08.09.2025 |
|---|
| Article acceptance date: | 24.10.2025 |
|---|
| Publication date: | 11.11.2025 |
|---|
| Publisher: | John Wiley & Sons Ltd. |
|---|
| Year of publishing: | 2025 |
|---|
| Number of pages: | 32 str. |
|---|
| Numbering: | Vol. 42, iss. 12, [article no.] e-70163 |
|---|
| PID: | 20.500.12556/DKUM-95960  |
|---|
| UDC: | 004.8 |
|---|
| ISSN on article: | 1468-0394 |
|---|
| COBISS.SI-ID: | 257158915 |
|---|
| DOI: | 10.1111/exsy.70163 |
|---|
| Copyright: | © 2025 The Author(s) |
|---|
| Publication date in DKUM: | 14.11.2025 |
|---|
| Views: | 348 |
|---|
| Downloads: | 9 |
|---|
| Metadata: |    |
|---|
| Categories: | Misc.
|
|---|
|
:
|
Copy citation |
|---|
| | | | Average score: | (0 votes) |
|---|
| Your score: | Voting is allowed only for logged in users. |
|---|
| Share: |  |
|---|
Hover the mouse pointer over a document title to show the abstract or click
on the title to get all document metadata. |
