Research map / Architectures and efficiency / Attention and transformers
MLP-3D: A MLP-Like 3D Architecture With Grouped Time Mixing
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use 224² rather than 128² input (ablation) | MLP-3D-XS with one path · Something-Something V2 val | top-1 accuracy: 64.6 -> 65.4 | Helped |
| Use 224² rather than 128² input (ablation) | MLP-3D-XS with one path · Kinetics-400 val | top-1 accuracy: 75.0 -> 76.5 | Helped |
| Add a second path with the input background removed by temporal mean subtraction (ablation) | MLP-3D-XS with 128² input · Something-Something V2 val | top-1 accuracy: 64.6 -> 65.6 | Helped |
| Add a second path with the input background removed by temporal mean subtraction (ablation) | MLP-3D-XS with 128² input · Kinetics-400 val | top-1 accuracy: 75.0 -> 76.2 | Helped |
| Increase network depth from MLP-3D-XS to MLP-3D-L (ablation) | MLP-3D, 224² input with two paths · Something-Something V2 val | top-1 accuracy: 66.0 -> 68.5 | Helped |
| Increase network depth from MLP-3D-XS to MLP-3D-L (ablation) | MLP-3D, 224² input with two paths · Kinetics-400 val | top-1 accuracy: 77.2 -> 81.3 | Helped |
| Use shift-token GTM rather than short-range or long-range GTM alone (ablation) | MLP-3D variants · Something-Something V2 | top-1 accuracy | Helped |
| Combine short-range GTM with long-range GTM or shift-window GTM (ablation) | MLP-3D variants · Something-Something V2 | top-1 accuracy | Helped |
| Increase GTM group size continuously (ablation) | MLP-3D variants · Something-Something V2 | top-1 accuracy | Mixed |
Closest papers by meaning
- MLP-Mixer: An all-MLP Architecture for Vision (NeurIPS 2021 · similarity 0.88)
- MiCT: Mixed 3D/2D Convolutional Tube for Human Action Recognition (CVPR 2018 · similarity 0.88)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.88)
- MAXIM: Multi-Axis MLP for Image Processing (CVPR 2022 · similarity 0.87)
- X3D: Expanding Architectures for Efficient Video Recognition (CVPR 2020 · similarity 0.87)
- AS-MLP: An Axial Shifted MLP Architecture for Vision (ICLR 2022 · 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.