Research map / Detection and segmentation / Open-set and few-shot detection
Synthetic Object Compositions for Scalable and Accurate Learning in Detection, Segmentation, and Grounding
Weikai Huang, Jieyu Zhang, Taoyang Jia, Chenhao Zheng, Ziqi Gao, Jae Sung Park, Ranjay Krishna
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 50K SOC images to O365+GoldG | MM-Grounding-DINO · LVIS AP | AP: 20.1 -> 29.8 | Improved |
| Add 20M GRIT vs 50K SOC | MM-Grounding-DINO · LVIS AP | AP: +7.0 (GRIT) vs +9.7 (SOC-50K) | Improved |
| Scale SOC from 100K to 400K | MM-Grounding-DINO · OdinW-35 | avg AP: 21.0 -> 22.8 | Improved |
| Add 50K SOC | MM-Grounding-DINO · OdinW-35 | avg AP: 20.3 -> 20.5 | No clear effect |
| Add 100K SOC to GRIT+V3Det | MM-Grounding-DINO · LVIS APrare | APrare: 23.6 -> 29.8 | Improved |
| Add V3Det to O365+GoldG | MM-Grounding-DINO · gRefCOCO no-target accuracy | Nacc: 89.3 -> 89.3 | No clear effect |
| Add SOC-FC-100K | MM-Grounding-DINO · gRefCOCO no-target accuracy | Nacc: 89.3 -> 97.7 | Improved |
| Add 50K SOC to LVIS training | APE · LVIS v1 val | APrare: 40.87 -> 44.70 | Improved |
| Add 3D geometric layout vs random 2D layout (ablation) | SOC pipeline · COCO instance seg | AP: 9.07 -> 10.03 | Improved |
| Add camera configuration augmentation (ablation) | SOC pipeline · COCO instance seg | AP: 10.03 -> 10.58 | Improved |
| Add generative harmonization (inpainting+relighting) (ablation) | SOC pipeline · COCO instance seg | AP: 6.28 -> 10.58 | Improved |
| Add mask-area-weighted blending (ablation) | SOC pipeline · COCO instance seg | AP: 10.58 -> 12.79 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- +GRIT (20M) — GRIT: Faster and Better Image Captioning Transformer Using Dual Visual Features (ECCV 2022) · better on 3 of 3 shared results
- X-Paste — X-Paste: Revisiting Scalable Copy-Paste for Instance Segmentation using CLIP and StableDiffusion (ICML 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DAViD: Data-efficient and Accurate Vision Models from Synthetic Data (ICCV 2025 · similarity 0.89)
- What Makes Synthetic Data Effective in Image Segmentation (ICML 2026 · similarity 0.89)
- Cut, Paste and Learn: Surprisingly Easy Synthesis for Instance Detection (ICCV 2017 · similarity 0.88)
- FreeMask: Synthetic Images with Dense Annotations Make Stronger Segmentation Models (NeurIPS 2023 · similarity 0.88)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.87)
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 2024 · similarity 0.86)
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.
