Research map / Detection and segmentation / Open-set and few-shot detection
Compositing-aware Image Search
Hengshuang Zhao, Xiaohui Shen, Zhe Lin, Kalyan Sunkavalli, Brian Price, Jiaya Jia
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 |
|---|---|---|---|
| extend triplets with top-10 retrieved foregrounds chosen by combined semantic and shape criteria (ablation) | Basic triplet training · Compositing evaluation set | mean MAP: 38.26 -> 45.70 | Improved |
| use shape-only triplet extension (ablation) | Basic triplet training · Compositing evaluation set | mean MAP: 38.26 -> 35.81 | Hurt |
| use semantic-context triplet extension (ablation) | Basic triplet training · Compositing evaluation set | mean MAP: 38.26 -> 44.18 | Improved |
| add shape to semantic triplet extension (ablation) | Semantics extension · Compositing evaluation set | mean MAP: 44.18 -> 45.70 | Improved |
| add MCB module with 20480 output dimension (ablation) | Shared network · Compositing evaluation set | mean MAP: 46.02 -> 48.42 | Improved |
| increase MCB dimension from 20480 to 40960 (ablation) | MCB · Compositing evaluation set | mean MAP: 48.42 -> 47.91 | No clear effect |
| use two-stage training instead of one-stage (ablation) | CAIS · Compositing evaluation set | mean MAP: 44.65 -> 48.42 | Improved |
| use MCB-based compact network instead of separate per-category modules (ablation) | CAIS · Compositing evaluation set | mean MAP: 48.69 -> 48.42 | No clear effect |
| use CAIS instead of classification-feature search | Classification feature · Compositing evaluation set | mean MAP: 36.78 -> 48.42 | Improved |
| use CAIS instead of classification feature in user study | Classification feature · User study | preference: 70.38 | Improved |
| shape feature on car category | Classification feature · Compositing evaluation set (car) | MAP: 14.93 -> 34.77 | Mixed |
Closest papers by meaning
- Interpretable Foreground Object Search As Knowledge Distillation (ECCV 2020 · similarity 0.90)
- GALA: Toward Geometry-and-Lighting-Aware Object Search for Compositing (ECCV 2022 · similarity 0.90)
- Unconstrained Foreground Object Search (ICCV 2019 · similarity 0.88)
- ICAR: Image-Based Complementary Auto Reasoning (AAAI 2024 · similarity 0.86)
- ORIDa: Object-centric Real-world Image Composition Dataset (CVPR 2025 · similarity 0.84)
- $\textit{Bifr\"ost}$: 3D-Aware Image Compositing with Language Instructions (NeurIPS 2024 · similarity 0.84)
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.
