Research map / Detection and segmentation / Open-set and few-shot detection
Learning Object Placement by Inpainting for Compositional Data Augmentation
Lingzhi Zhang, Tarmily Wen, Jie Min, Jiancong Wang, David Han, Jianbo Shi
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 diversity loss to the cVAE-GAN placement model (cVAE-GAN+Div) | cVAE-GAN · Object placement diversity | diversity: 0.0219 -> 0.0335 | Improved |
| use PlaceNet pairwise-distance diversity loss with spectral-norm discriminator instead of cVAE-GAN+Div | placement model · Object placement | FID: 53.54 -> 48.15 | Improved |
| use PlaceNet placements instead of random placement for augmentation | YOLOv3 on KITTI · KITTI object detection | overall mAP: 0.203 -> 0.371 | Improved |
| augment training data with random placement | YOLOv3 on KITTI · KITTI object detection | overall mAP: 0.359 -> 0.203 | Hurt |
| augment with PlaceNet object placement | YOLOv3 · Cityscapes object detection | overall mAP: 0.265 -> 0.302 | Improved |
| augment with PlaceNet object placement | YOLOv3 · KITTI object detection | overall mAP: 0.359 -> 0.371 | Improved |
| augment with PlaceNet object placement | YOLOv3 · Cityscapes object detection, car class | mAP: 0.754 -> 0.749 | Hurt |
| augment with PlaceNet object placement | YOLOv3 · Cityscapes object detection, rider class | mAP: 0.238 -> 0.381 | Improved |
| augment with PlaceNet object placement | Mask R-CNN · Cityscapes instance segmentation | overall mAP: 0.286 -> 0.302 | Improved |
| augment with PlaceNet object placement | Mask R-CNN · KITTI instance segmentation | overall mAP: 0.235 -> 0.307 | Improved |
| augment with random placement | Mask R-CNN · Cityscapes instance segmentation | overall mAP: 0.286 -> 0.279 | Hurt |
| use diverse PlaceNet placements instead of less diverse cVAE-GAN | YOLOv3 on KITTI · KITTI object detection | overall mAP: 0.274 -> 0.371 | Improved |
Closest papers by meaning
- TopNet: Transformer-Based Object Placement Network for Image Compositing (CVPR 2023 · similarity 0.88)
- Seeing the Unseen: Visual Common Sense for Semantic Placement (CVPR 2024 · similarity 0.88)
- Context-aware Synthesis and Placement of Object Instances (NeurIPS 2018 · similarity 0.88)
- BOOTPLACE: Bootstrapped Object Placement with Detection Transformers (CVPR 2025 · similarity 0.87)
- Learning to Compose: Improving Object Centric Learning by Injecting Compositionality (ICLR 2024 · similarity 0.86)
- MonoPlace3D: Learning 3D-Aware Object Placement for 3D Monocular Detection (CVPR 2025 · 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.
