Research map / 3D vision / Human pose estimation
I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image
Gyeongsik Moon, Kyoung Mu Lee
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 lixel heatmap target (with uncertainty) over direct coordinate regression (ablation) | MeshNet · Human3.6M | MPJPE: 114.3 -> 92.6 | Improved |
| add spatial relationship preservation (fully-convolutional heatmaps) (ablation) | lixel heatmap MeshNet · Human3.6M | MPJPE: 92.6 -> 86.2 | Improved |
| use lixel heatmap instead of SMPL parameter regression (ablation) | MeshNet · Human3.6M | MPJPE: 100.3 -> 86.2 | Improved |
| increase lixel heatmap resolution from 8 to 64 (ablation) | MeshNet · Human3.6M | MPJPE: 100.2 -> 94.8 -> 86.2 | Improved |
| use lixel heatmap instead of voxel heatmap at 8 resolution (ablation) | MeshNet · Human3.6M | GPU memory: voxel 4.3 GB at 8x8x8 vs lixel 3.4 GB at 8,8,8 | Improved |
| add PoseNet cascaded before MeshNet (ablation) | I2L-MeshNet · Human3.6M | MPJPE: 86.2 -> 81.8 | Improved |
| feed ground-truth 3D pose to MeshNet (ablation) | I2L-MeshNet · Human3.6M | MPJPE: 86.2 -> 25.5 | Improved |
| marginalize at the last upsampling output instead of on ResNet output (ablation) | MeshNet · Human3.6M | MPJPE: 93.5 -> 86.2 | Improved |
| use average pooling for marginalization instead of max pooling (ablation) | MeshNet · Human3.6M | MPJPE: 93.5 -> 86.2 | Improved |
| use average pooling instead of learned weighted sum (ablation) | MeshNet · Human3.6M | MPJPE: 89.4 -> 86.2 | Improved |
| use lixel 1D heatmap instead of 2.5D heatmap (ablation) | MeshNet · Human3.6M | MPJPE: 100.4 -> 94.8 at 32 resolution | Improved |
| add MeshNet pose loss L_pose (ablation) | I2L-MeshNet · Human3.6M | MPJPE: 84.5 -> 81.8 | Improved |
Later papers that compared with this method
12 later accepted papers report I2L-MeshNet as a baseline in their results tables; 12 of them report a better number on most of the results they share with it.
- 3D Hand Sequence Recovery from Real Blurry Images and Event Stream (ECCV 2024 · better on 2 of 2 shared results)
- HandDGP: Camera-Space Hand Mesh Prediction with Differentiable Global Positioning (ECCV 2024 · better on 1 of 1 shared result)
- Recovering 3D Hand Mesh Sequence From a Single Blurry Image: A New Dataset and Temporal Unfolding (CVPR 2023 · better on 2 of 2 shared results)
- HandOccNet: Occlusion-Robust 3D Hand Mesh Estimation Network (CVPR 2022 · better on 1 of 1 shared result)
- Learning Visibility for Robust Dense Human Body Estimation (ECCV 2022 · better on 1 of 1 shared result)
- Towards Accurate Alignment in Real-Time 3D Hand-Mesh Reconstruction (ICCV 2021 · better on 8 of 8 shared results)
- End-to-End Human Pose and Mesh Reconstruction with Transformers (CVPR 2021 · better on 5 of 5 shared results)
- Camera-Space Hand Mesh Recovery via Semantic Aggregation and Adaptive 2D-1D Registration (CVPR 2021 · better on 3 of 4 shared results)
4 more not listed.
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
- Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose (ECCV 2020 · similarity 0.88)
- Convolutional Mesh Regression for Single-Image Human Shape Reconstruction (CVPR 2019 · similarity 0.87)
- Co-Evolution of Pose and Mesh for 3D Human Body Estimation from Video (ICCV 2023 · similarity 0.87)
- V2V-PoseNet: Voxel-to-Voxel Prediction Network for Accurate 3D Hand and Human Pose Estimation From a Single Depth Map (CVPR 2018 · similarity 0.86)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.86)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 2024 · 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.
