Research map / 3D vision / Depth and stereo
AGG-Net: Attention Guided Gated-Convolutional Network for Depth Image Completion
Dongyue Chen, Tingxuan Huang, Zhimin Song, Shizhuo Deng, Tong Jia
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 color features via concatenation to depth-only baseline (ablation) | UNet-like depth completion · NYU-Depth V2 | RMSE: 0.136 -> 0.122 | Improved |
| replace vanilla convolution with gated convolution and De-GConv (ablation) | concat-fusion UNet · NYU-Depth V2 | RMSE: 0.122 -> 0.115 | Improved |
| replace GConv with AG-GConv (contextual attention from depth and color) (ablation) | AGG-Net · NYU-Depth V2 | RMSE: 0.115 -> 0.105 | Improved |
| add AG-SC module on color skip connection (ablation) | GConv concat model · NYU-Depth V2 | RMSE: 0.115 -> 0.103 | Improved |
| add AG-SC on top of AG-GConv (ablation) | AGG-Net · NYU-Depth V2 | RMSE: 0.105 -> 0.092 | Improved |
| add pre-filling network (ablation) | AGG-Net · NYU-Depth V2 | RMSE: 0.105 -> 0.092 | Improved |
| replace MSE with Huber loss (ablation) | AGG-Net · NYU-Depth V2 | RMSE: 0.104 -> 0.100 | Improved |
| add edge persistence loss to Huber loss (ablation) | AGG-Net · NYU-Depth V2 | RMSE: 0.100 -> 0.092 | Improved |
| increase encoder layers beyond m=4 | AGG-Net · NYU-Depth V2 | RMSE | No clear effect |
| use kernel size larger than 3 | AGG-Net · NYU-Depth V2 | RMSE | Hurt |
| increase CA hidden ratio beyond r=4 | AGG-Net · NYU-Depth V2 | RMSE | No clear effect |
Later papers that compared with this method
1 later accepted paper reports AGG-Net as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Completion as Enhancement: A Degradation-Aware Selective Image Guided Network for Depth Completion (CVPR 2025 · better on 1 of 1 shared result)
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
- CostDCNet: Cost Volume Based Depth Completion for a Single RGB-D Image (ECCV 2022 · similarity 0.89)
- Deep Depth Completion of a Single RGB-D Image (CVPR 2018 · similarity 0.89)
- RigNet++: Semantic Assisted Repetitive Image Guided Network for Depth Completion (IJCV 2025 · similarity 0.88)
- Learning Joint 2D-3D Representations for Depth Completion (ICCV 2019 · similarity 0.88)
- Efficient Indoor Depth Completion Network Using Mask-adaptive Gated Convolution (AAAI 2025 · similarity 0.87)
- Completion as Enhancement: A Degradation-Aware Selective Image Guided Network for Depth Completion (CVPR 2025 · similarity 0.87)
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.
