Research map / Detection and segmentation / Weakly supervised detection and segmentation
BigDatasetGAN: Synthesizing ImageNet With Pixel-Wise Annotations
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 BigGAN-sim training data (ablation) | Sup.IN · ImageNet pixel-wise benchmark, mean | mIoU: 55.7 -> 59.0 | Helped |
| Use online rather than offline BigGAN sampling (ablation) | Sup.IN + BigGAN · ImageNet pixel-wise benchmark, mean | mIoU: 59.0 -> 60.4 | Helped |
| Add BigGAN-sim training data (ablation) | SupSelfCon · ImageNet pixel-wise benchmark, mean | mIoU: 57.8 -> 61.6 | Helped |
| Add VQGAN-sim training data (ablation) | SupSelfCon · ImageNet pixel-wise benchmark, mean | mIoU: 57.8 -> 61.6 | Helped |
| Add BigGAN-sim training data (ablation) | DenseCL · ImageNet pixel-wise benchmark, MC-16 | mIoU: 47.9 -> 58.6 | Helped |
| Add VQGAN-sim training data (ablation) | DenseCL · ImageNet pixel-wise benchmark, MC-16 | mIoU: 47.9 -> 59.3 | Helped |
| Add VQGAN-sim training data (ablation) | MoCo-v3 · ImageNet pixel-wise benchmark, mean | mIoU: 52.5 -> 62.7 | Helped |
| Increase BigGAN-sim dataset size from 22k to 220k (ablation) | ImageNet segmentation model · ImageNet pixel-wise benchmark | mIoU: 7 points | Helped |
| Pre-train with BigGAN-sim supervision and contrastive learning | DenseCL backbone · MS-COCO val'17, 1× schedule | APbb: +0.4 APbb | Helped |
| Pre-train with BigGAN-sim supervision and contrastive learning | DenseCL backbone · MS-COCO val'17, 1× schedule | APmk: +0.3 APmk | Helped |
| Pre-train with BigGAN-sim supervision and contrastive learning | DenseCL backbone · PASCAL VOC val2012 | mIoU: 0.5% | Helped |
| Pre-train with BigGAN-sim supervision and contrastive learning | DenseCL backbone · Cityscapes val | mIoU: 75.7 -> 75.6 | No clear effect |
Closest papers by meaning
- DatasetGAN: Efficient Labeled Data Factory With Minimal Human Effort (CVPR 2021 · similarity 0.92)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.89)
- Large Scale Adversarial Representation Learning (NeurIPS 2019 · similarity 0.89)
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.88)
- Scaling Up GANs for Text-to-Image Synthesis (CVPR 2023 · similarity 0.88)
- Training Generative Adversarial Networks from Incomplete Observations using Factorised Discriminators (ICLR 2020 · similarity 0.88)
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.