Research map / Detection and segmentation / CNN features
Oriented Response Networks
Yanzhao Zhou, Qixiang Ye, Qiang Qiu, Jianbin Jiao
Check 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 |
|---|---|---|---|
| Replace standard convolutions with ORConv using ARFs and ORAlign | CNN · MNIST-original | error rate (%): 0.73 -> 0.57 | Improved |
| Replace standard convolutions with ARFs and ORPooling | CNN · MNIST-rot | error rate (%): 2.82 -> 1.37 | Improved |
| Use ARFs with ORAlign | CNN · MNIST-rot+ | error rate (%): 2.19 -> 1.12 | Improved |
| Train with ARFs and ORAlign instead of a conventional CNN | CNN · MNIST-original to MNIST-rot | error rate (%): 56.28 -> 16.21 | Improved |
| Add ORAlign to the OR-STN localization network (ablation) | OR-STN · MNIST-half-rot | classification error (%): 2.43 -> 3.61 | Hurt |
| Upgrade the STN localization network from CNN to ORN without alignment | STN · MNIST-half-rot | orientation standard deviation: 0.745 -> 0.397 | Improved |
| Upgrade the STN localization network from CNN to ORN without alignment | STN · MNIST-half-rot | classification error (%): 3.38 -> 2.43 | Improved |
| Replace the CNN with ORN-8(ORAlign) | CNN · MNIST-rot-12k | error rate (%): 4.34 -> 2.25 | Improved |
| Replace the CNN branches in TIPooling with ORN-8 | TIPooling · MNIST-rot-12k | error rate (%): 1.93 -> 1.54 | Improved |
| Replace standard convolutions with ORConv | VGG · CIFAR10 | classification error (%): 6.32 -> 5.47 | Improved |
| Replace standard convolutions with ORConv | VGG · CIFAR100 | classification error (%): 28.49 -> 27.03 | Improved |
| Replace standard convolutions with ORConv | ResNet · CIFAR10 | classification error (%): 6.43 -> 5.31 | Improved |
Closest papers by meaning
- Rotation Equivariant Vector Field Networks (ICCV 2017 · similarity 0.87)
- Learning Steerable Filters for Rotation Equivariant CNNs (CVPR 2018 · similarity 0.87)
- Harmonic Networks: Deep Translation and Rotation Equivariance (CVPR 2017 · similarity 0.87)
- RRL: Regional Rotate Layer in Convolutional Neural Networks (AAAI 2022 · similarity 0.87)
- RotDCF: Decomposition of Convolutional Filters for Rotation-Equivariant Deep Networks (ICLR 2019 · similarity 0.86)
- 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.
