Research map / Detection and segmentation / Open-set and few-shot detection
Generative Modeling for Small-Data Object Detection
Lanlan Liu, Michael Muelly, Jia Deng, Tomas Pfister, Li-Jia Li
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 synthetic images from DetectorGAN to real data | detector · NIH nodules (new annotations) | Nodule AP: 0.124 -> 0.236 | Improved |
| add synthetic images from DetectorGAN to real data | detector · NIH nodules (new annotations) | Nodule recall: 0.184 -> 0.649 | Improved |
| add unrolling step (ablation) | DetectorGAN · NIH nodules (new annotations) | Nodule AP: 0.203 -> 0.236 | Improved |
| maximize detection loss on synthetic images instead of minimizing it (ACGAN-like) (ablation) | DetectorGAN · NIH nodules (new annotations) | Nodule AP: 0.154 -> 0.236 | Improved |
| add ACGAN-like detection loss on synthetic images (ablation) | CycleGAN-style generator · NIH nodules (new annotations) | Nodule AP: 0.196 -> 0.154 | Hurt |
| use DetectorGAN instead of CycleGAN + BboxLoss | synthetic data source · NIH nodules (new annotations) | Nodule AP: 0.196 -> 0.236 | Improved |
| add DetectorGAN synthetic images to RetinaNet | RetinaNet · NIH nodules (old annotations) | average localization accuracy over T_IOU: 0.09 -> 0.23 | Improved |
| add DetectorGAN synthetic images to RetinaNet | RetinaNet · NIH nodules (old annotations) | average localization accuracy over T_IOBB: 0.09 -> 0.22 | Improved |
| add DetectorGAN synthetic images | detector · Cityscapes pedestrian detection | AP: 0.593 -> 0.613 | Improved |
| add detector in the loop to PS-GAN | PS-GAN synthetic data · Cityscapes pedestrian detection | AP: 0.602 -> 0.613 | Improved |
| add pix2pix synthetic images | detector · Cityscapes pedestrian detection | AP: 0.593 -> 0.574 | Hurt |
| use DetectorGAN instead of CycleGAN + BboxLoss images | radiologist rating · NIH user study (object) | mean Likert: 1.31 -> 2.69 | Improved |
Closest papers by meaning
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 2024 · similarity 0.87)
- Generative Adversarial Learning Towards Fast Weakly Supervised Detection (CVPR 2018 · similarity 0.86)
- Believing is Seeing: Unobserved Object Detection using Generative Models (CVPR 2025 · similarity 0.86)
- InstaGen: Enhancing Object Detection by Training on Synthetic Dataset (CVPR 2024 · similarity 0.85)
- Intriguing Properties of Generative Classifiers (ICLR 2024 · similarity 0.85)
- ODGEN: Domain-specific Object Detection Data Generation with Diffusion Models (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.
