Research map / 3D vision / Human pose estimation
FAFA: Frequency-Aware Flow-Aided Self-Supervision for Underwater Object Pose Estimation
Jingyi Tang, Gu Wang, Zeyu Chen, Shengquan Li, Xiu Li, Xiangyang Ji
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 frequency-aware FFT augmentation (amplitude mix and dropout) in synthetic pre-training (ablation) | FAFA pre-training · DeepURL | ADD 0.1d: 50.18 -> 73.36 | Improved |
| add amplitude dropout during pre-training (ablation) | FAFA pre-training · DeepURL | ADD 0.1d: 55.03 -> 73.36 | Improved |
| add amplitude mix during pre-training (ablation) | FAFA pre-training · DeepURL | ADD 0.1d: 58.54 -> 73.36 | Improved |
| apply FFT augmentation also on real images in self-supervision (ablation) | FAFA self-supervision · DeepURL | MEAN: 73.46 -> 67.71 | Hurt |
| add photometric image-level loss L_photo (ablation) | FAFA self-supervision · DeepURL | 5 deg: 50.31 -> 62.02 | Improved |
| add warp-mask loss L_warp-mask (ablation) | FAFA self-supervision · DeepURL | 5 cm: 43.90 -> 49.36 | Improved |
| add feature-level alignment on top of image-level losses (ablation) | FAFA self-supervision · DeepURL | MEAN: 70.68 -> 73.46 | Improved |
| add self-supervised training on unlabeled real images | FAFA pre-trained · ROV6D Pool-Occluded | MEAN: 85.06 -> 91.39 | Improved |
| use FAFA self-supervision instead of PFC | PFC · ROV6D Pool-Occluded | 5 cm: 72.10 -> 86.71 | Improved |
| use FAFA instead of PFC on DeepURL | PFC · DeepURL | ADD 0.1d: 98.87 -> 97.64 | Hurt |
| use FAFA instead of PFC on DeepURL | PFC · DeepURL | 5 cm: 44.10 -> 50.92 | Improved |
| add feature-level loss alone without image-level losses (ablation) | FAFA self-supervision · DeepURL | ADD 0.1d: 97.00 -> 95.73 | Hurt |
Closest papers by meaning
- Perspective Flow Aggregation for Data-Limited 6D Object Pose Estimation (ECCV 2022 · similarity 0.86)
- WebUOT-1M: Advancing Deep Underwater Object Tracking with A Million-Scale Benchmark (NeurIPS 2024 · similarity 0.85)
- Wavelet-based Fourier Information Interaction with Frequency Diffusion Adjustment for Underwater Image Restoration (CVPR 2024 · similarity 0.83)
- Diving into Underwater: Segment Anything Model Guided Underwater Salient Instance Segmentation and A Large-scale Dataset (ICML 2024 · similarity 0.83)
- Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · similarity 0.83)
- Adaptive Dual-domain Learning for Underwater Image Enhancement (AAAI 2025 · 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.
