Research map / Graph learning / Graph algorithms and clustering
Learning Structural Similarity of User Interface Layouts using Graph Networks
Dipu Manandhar, Dan Ruta, John Collomosse
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add triplet supervision with IoU weak labels (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 48.3 -> 51.3 | Improved |
| add triplet supervision with IoU weak labels (ablation) | GCN-CNN layout embedding · RICO | MPixAcc@10 (%): 56.5 -> 61.0 | Improved |
| add containment feature (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 47.4 -> 48.0 | Improved |
| add directed graph connectivity (ablation) | GCN-CNN layout embedding · RICO | MPixAcc@10 (%): 57.0 -> 57.7 | Improved |
| add self-attention readout pooling (ablation) | GCN-CNN layout embedding · RICO | MPixAcc@10 (%): 57.7 -> 58.0 | Improved |
| add triplet supervision at the end of cumulative stack (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 48.3 -> 51.3 | Improved |
| use strided convolution instead of upsampling in decoder (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 47.1 -> 48.3 | Improved |
| use directed instead of undirected graph, triplet trained (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 51.0 -> 51.3 | Improved |
| add containment feature with triplet supervision (ablation) | GCN-CNN layout embedding · RICO | MIoU@5 (%): 53.4 -> 54.1 | Improved |
| add containment feature (ablation) | GCN-CNN layout embedding · RICO | MIoU@1 (%): 61.9 -> 60.0 | Hurt |
| use self-attention readout instead of average pooling (512-D, unsupervised) (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 47.4 -> 47.1 | No clear effect |
| use self-attention readout instead of average/inverse pooling (2048-D, TRI) (ablation) | GCN-CNN layout embedding · RICO | MIoU@10 (%): 51.1 -> 51.3 | No clear effect |
Closest papers by meaning
- DeepLayout: Learning Neural Representations of Circuit Placement Layout (ICML 2025 · similarity 0.85)
- Neural Architecture Retrieval (ICLR 2024 · similarity 0.81)
- MaskPLAN: Masked Generative Layout Planning from Partial Input (CVPR 2024 · similarity 0.81)
- LayoutNUWA: Revealing the Hidden Layout Expertise of Large Language Models (ICLR 2024 · similarity 0.81)
- AesthetiQ: Enhancing Graphic Layout Design via Aesthetic-Aware Preference Alignment of Multi-modal Large Language Models (CVPR 2025 · similarity 0.81)
- No More Ambiguity in 360deg Room Layout via Bi-Layout Estimation (CVPR 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
