Research map / Architectures and efficiency / CNN design
Kernel Transformer Networks for Compact Spherical Convolution
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 KTN instead of SphConv | Spherical convolution of the MNIST CNN · Spherical MNIST test | accuracy: 98.72 -> 97.94 | Hurt |
| Use KTN instead of SphConv | Faster R-CNN VGG convolution · Pano2Vid test | RMSE: 1.50 -> 1.53 | Hurt |
| Use KTN instead of SphConv | Faster R-CNN VGG object detection · Pascal VOC 2007 validation | accuracy: 63.54 -> 69.48 | Helped |
| Add the 1x1 pointwise convolution that captures cross-channel correlation (ablation) | KTN · Spherical convolution accuracy | accuracy | Helped |
| Replace the learned kernel transformation with tangent-plane-to-sphere projection | KTN · Pano2Vid test | RMSE: 1.53 -> 4.24 | Hurt |
| Share the learned transformation across source CNNs with the same architecture | KTN · Pano2Vid | RMSE | No clear effect |
| Use a learned kernel transformation instead of feature projection | Spherical convolution on deep CNNs · Pano2Vid | RMSE: 2.46 -> 1.53 | Helped |
Closest papers by meaning
- Learning Spherical Convolution for Fast Features from 360° Imagery (NeurIPS 2017 · similarity 0.89)
- Kervolutional Neural Networks (CVPR 2019 · similarity 0.86)
- Are Large Kernels Better Teachers than Transformers for ConvNets? (ICML 2023 · similarity 0.86)
- Scaling Up Your Kernels to 31x31: Revisiting Large Kernel Design in CNNs (CVPR 2022 · similarity 0.86)
- Spherical CNNs on Unstructured Grids (ICLR 2019 · similarity 0.86)
- KPConvX: Modernizing Kernel Point Convolution with Kernel Attention (CVPR 2024 · similarity 0.84)
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.