Research map / Image generation / Image restoration
Deep Symmetric Network for Underexposed Image Enhancement With Recurrent Attentional Learning
Lin Zhao, Shao-Ping Lu, Tao Chen, Zhenglu Yang, Ariel Shamir
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 backward loss (two-way propagation) (ablation) | Deep Symmetric Network · LOL | PSNR change: -1.28 | Improved |
| add recurrent learning (ablation) | Deep Symmetric Network · LOL | PSNR change: -1.08 | Improved |
| add attention in RRAM (ablation) | Deep Symmetric Network · LOL | PSNR change: -3.10 | Improved |
| use RRAM instead of plain residual blocks (ablation) | Deep Symmetric Network · LOL | PSNR change: -3.50 | Improved |
| use RRAM instead of SE residual blocks (ablation) | Deep Symmetric Network · LOL | PSNR change: -1.32 | Improved |
| use RRAM instead of DenseNet (ablation) | Deep Symmetric Network · LOL | PSNR change: -2.74 | Improved |
| use three recurrence rounds instead of two (ablation) | Deep Symmetric Network · LOL | PSNR change: -0.42 | Improved |
| use 8 invertible blocks instead of 4 (ablation) | Deep Symmetric Network · LOL | PSNR change: -1.42 | Improved |
| use pretrained VGG encoder-decoder instead of Haar wavelets (ablation) | Deep Symmetric Network · LOL | PSNR change: -1.74 | Improved |
| use L2 loss instead of L1 (ablation) | Deep Symmetric Network · LOL | PSNR change: -0.98 | Improved |
| use L2 loss instead of PSNR loss (ablation) | Deep Symmetric Network · LOL | PSNR change: -0.20 | Improved |
| add recurrent learning (ablation) | Deep Symmetric Network · MIT-Adobe FiveK | PSNR: 23.32/0.888 -> 24.27/0.900 | Improved |
Closest papers by meaning
- Underexposed Photo Enhancement Using Deep Illumination Estimation (CVPR 2019 · similarity 0.87)
- Image Restoration Using Very Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections (NeurIPS 2016 · similarity 0.85)
- BNUDC: A Two-Branched Deep Neural Network for Restoring Images From Under-Display Cameras (CVPR 2022 · similarity 0.85)
- Learning Multi-Scale Photo Exposure Correction (CVPR 2021 · similarity 0.85)
- VideoOneNet: Bidirectional Convolutional Recurrent OneNet with Trainable Data Steps for Video Processing (ICML 2020 · similarity 0.85)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · 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.
