Research map / Graph learning / Knowledge and temporal graphs
Learning Object Placement via Dual-Path Graph Completion
Siyuan Zhou, Liu Liu, Li Niu, Liqing Zhang
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 |
|---|---|---|---|
| use GracoNet instead of PlaceNet | object placement · OPA test (positive samples) | accuracy: 0.683 -> 0.847 | Improved |
| use GracoNet instead of PlaceNet | object placement · OPA test (positive samples) | FID: 36.69 -> 27.75 | Improved |
| add NEH multi-scale nodes instead of global vectors (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.793 -> 0.828 | Improved |
| add PSN cross-attention instead of concat+fc (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.828 -> 0.847 | Improved |
| add small-scale background nodes (n=[2,4,8] vs [2]) (ablation) | GracoNet · OPA test (positive samples) | FID/LPIPS: LPIPS 0.136 -> 0.206 | Improved |
| extend background head with 16x16 nodes (n=[2,4,8,16]) (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.847 -> 0.837 | Hurt |
| add placement encodings p^K and p^V (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.793 -> 0.847 | Improved |
| make placement encodings independent across attention heads (ablation) | GracoNet · OPA test (positive samples) | LPIPS: 0.170 -> 0.206 | Improved |
| add classification loss L_cls_s (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.734 -> 0.847 | Improved |
| add reconstruction loss L_rec_s (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.767 -> 0.847; FID 25.53 -> 27.75 | Mixed |
| add KL loss L_kld_s (ablation) | GracoNet · OPA test (positive samples) | accuracy: 0.844 -> 0.847 | Improved |
| add unsupervised adversarial loss L_adv_u (ablation) | GracoNet · OPA test (positive samples) | LPIPS: 0.038 -> 0.206 | Improved |
Closest papers by meaning
- TopNet: Transformer-Based Object Placement Network for Image Compositing (CVPR 2023 · similarity 0.87)
- BOOTPLACE: Bootstrapped Object Placement with Detection Transformers (CVPR 2025 · similarity 0.84)
- Point-based Instance Completion with Scene Constraints (ICLR 2025 · similarity 0.82)
- Multi-Scene Generalized Trajectory Global Graph Solver with Composite Nodes for Multiple Object Tracking (AAAI 2024 · similarity 0.82)
- Completing Visual Objects via Bridging Generation and Segmentation (ICML 2024 · similarity 0.82)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.82)
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.
