Research map / Detection and segmentation / Weakly supervised detection and segmentation
Learning from Synthetic Data via Provenance-Based Input Gradient Guidance
Koshiro Nagano, Ryo Fujii, Ryo Hachiuma, Fumiaki Sato, Taiki Sekii, Hideo Saito
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 provenance-based input gradient guidance loss | CAM + CutMix (VGG16) · CUB WSOL | MaxBoxAccV2 mean (%): 62.3 -> 65.1 | Improved |
| add provenance loss | CAM + ResizeMix · CUB WSOL | MaxBoxAccV2 mean (%): 57.6 -> 62.2 | Improved |
| add provenance loss | CAM + PuzzleMix · CUB WSOL | MaxBoxAccV2 mean (%): 56.8 -> 58.2 | Improved |
| add provenance loss | CAM + CutMix (VGG16) · CUB WSOL at IoU 0.7 | MaxBoxAccV2 (%): 28.6 -> 23.1 | Hurt |
| add provenance loss | SAT + CutMix (DeiT-S) · CUB WSOL | MaxBoxAccV2 mean (%): 91.5 -> 92.1 | Improved |
| add CutMix augmentation | SAT (DeiT-S) · CUB WSOL | MaxBoxAccV2 mean (%): 91.4 -> 91.5 | No clear effect |
| add provenance loss to skeleton BatchMix | SKP · UCF101-24 | AP@0.5 (%): 38.0 -> 39.7 | Improved |
| add provenance loss | ALIA · Waterbirds | Top-1 accuracy (%): 71.4 -> 82.3 | Improved |
| add provenance loss | ALIA · CUB | Top-1 accuracy (%): 71.7 -> 72.0 | Improved |
| add provenance loss | ALIA · iWildCam | Top-1 accuracy (%): 83.5 -> 84.4 | Improved |
| use provenance mask instead of random mask for gradient regularization (ablation) | CutMix + provenance loss · CUB WSOL | MaxBoxAccV2 mean (%): 60.5 -> 65.1 | Improved |
| use provenance mask instead of unmasked gradient regularization (ablation) | CutMix + provenance loss · CUB WSOL | MaxBoxAccV2 mean (%): 61.1 -> 65.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CutMix — CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features (ICCV 2019) · better on 4 of 4 shared results
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
- Provably Improving Generalization of Few-shot models with Synthetic Data (ICML 2025 · similarity 0.85)
- Training Data Provenance Verification: Did Your Model Use Synthetic Data from My Generative Model for Training? (CVPR 2025 · similarity 0.84)
- Real-Fake: Effective Training Data Synthesis Through Distribution Matching (ICLR 2024 · similarity 0.84)
- Less Supervision, Better Generalization: Weakly Supervised Fake Region Localization in Diffusion-Edited Images (NeurIPS 2026 · similarity 0.84)
- The Unmet Promise of Synthetic Training Images: Using Retrieved Real Images Performs Better (NeurIPS 2024 · similarity 0.83)
- Discriminative Score Function: Turning Pretrained Models into Functional Generative Priors (NeurIPS 2026 · similarity 0.83)
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.
