Research map / Image generation / Text-to-image diffusion
TMI: Text-to-Image Meets Image-to-Image for Complementary Data Synthesis to Boost Long-Tailed Instance Segmentation
Hyeonseop Song, Seokhun Choi, Hoseok Do
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 |
|---|---|---|---|
| combine T2I and I2I synthesis (TMI) vs real-only | LVIS Mask R-CNN style detector · LVIS box AP, ResNet-50 | AP box: 34.5 -> 38.1 | Improved |
| TMI vs real-only on rare classes | detector · LVIS rare, ResNet-50 | AP box rare: 24.0 -> 33.9 | Improved |
| MosaicFusion T2I-only on ResNet-50 | detector · LVIS box AP, ResNet-50 | AP box: 34.5 -> 34.1 | No clear effect |
| text-consistent filtering instead of high-threshold filtering of T2I pseudo-labels (ablation) | T2I branch · LVIS box AP, ResNet-50 | AP box: 35.8 -> 36.7 | Improved |
| rare-target copy-paste (DiverGen) vs real-only (ablation) | detector · LVIS rare, Swin-L | AP box rare: 41.4 -> 34.5 | Hurt |
| VRAIN I2I data only vs real-only (ablation) | detector · LVIS rare, Swin-L | AP box rare: 41.4 -> 42.9 | Improved |
| add VLM verification to VRAIN (ablation) | VRAIN · LVIS rare, Swin-L | AP box rare: 39.5 -> 42.9 | Improved |
| VLM verifier instead of CLIP crop for false positives (ablation) | VRAIN · verification benchmark | FP rate: 0.187 -> 0.033 | Improved |
| add EMA teacher labels on top of offline labels (ablation) | T2I branch · LVIS box AP, Swin-L | AP box: 49.4 -> 50.3 | Improved |
| use progressive instead of frozen teacher pseudo labels (ablation) | T2I branch · LVIS rare, Swin-L | AP box rare: 45.5 -> 47.9 | Improved |
| treat prompt-inconsistent predictions as unlabeled rather than discard or label (ablation) | T2I branch · LVIS box AP, Swin-L | AP box: 47.3 -> 49.4 | Improved |
| add I2I data to T2I-only training (ablation) | TMI · LVIS rare, Swin-L | AP box rare: 47.5 -> 49.1 | Improved |
Closest papers by meaning
- Iterative Self-Training with Class-Aware Text-to-Image Synthesis for Visual Task Learning (AAAI 2025 · similarity 0.87)
- MARS: Mixture of Auto-Regressive Models for Fine-grained Text-to-image Synthesis (AAAI 2025 · similarity 0.87)
- DINOv2 Meets Text: A Unified Framework for Image- and Pixel-Level Vision-Language Alignment (CVPR 2025 · similarity 0.86)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.86)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
- Leveraging Unpaired Data for Vision-Language Generative Models via Cycle Consistency (ICLR 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.
