Research map / Label-efficient and robust learning / Self-supervised and contrastive
SimFIR: A Simple Framework for Fisheye Image Rectification with Self-supervised Representation Learning
Hao Feng, Wendi Wang, Jiajun Deng, Wengang Zhou, Li Li, Houqiang 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 unified contrastive + position pretext pre-training (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 20.33 -> 22.47 | Improved |
| add contrastive loss only for pre-training (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 20.33 -> 20.70 | Improved |
| add position loss only for pre-training (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 20.33 -> 21.30 | Improved |
| add patch shuffling during pre-training (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 20.28 -> 22.47 | Improved |
| unfreeze encoder during fine-tuning instead of freezing (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 21.12 -> 22.47 | Improved |
| predict warping flow instead of distortion parameters or the rectified image (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 21.09 -> 22.47 (vs parameter); 20.50 -> 22.47 (vs image) | Improved |
| reduce patch size from 16 to 8 (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 22.47 -> 25.82 | Improved |
| increase patch size from 16 to 64 (ablation) | SimFIR ViT · Synthetic fisheye test set | PSNR: 22.47 -> 18.65 | Hurt |
| add boundary refinement module (BRM) (ablation) | SimFIR · Synthetic fisheye test set | PSNR: 18.01 -> 22.47 | Improved |
| use learnable upsampling flow prediction module instead of bilinear upsampling (ablation) | SimFIR · Synthetic fisheye test set | PSNR: 22.11 -> 22.47 | Improved |
Closest papers by meaning
- QueryCDR: Query-based Controllable Distortion Rectification Network for Fisheye Images (ECCV 2024 · similarity 0.91)
- Progressively Complementary Network for Fisheye Image Rectification Using Appearance Flow (CVPR 2021 · similarity 0.90)
- Learning to Calibrate Straight Lines for Fisheye Image Rectification (CVPR 2019 · similarity 0.89)
- FishEyeRecNet: A Multi-Context Collaborative Deep Network for Fisheye Image Rectification (ECCV 2018 · similarity 0.89)
- Innovating Real Fisheye Image Correction with Dual Diffusion Architecture (ICCV 2023 · similarity 0.86)
- Fisheye3R: Adapting Unified 3D Feed-Forward Foundation Models to Fisheye Lenses (ECCV 2026 · 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.
