https://github.com/ruoqizzz/Scale-invariant-CNNs

Study the Scale Invariance or Equivariance Convolutional Neural Network
https://github.com/ruoqizzz/Scale-invariant-CNNs

Last synced: about 1 month ago
JSON representation

Study the Scale Invariance or Equivariance Convolutional Neural Network

• Host: GitHub
• URL: https://github.com/ruoqizzz/Scale-invariant-CNNs
• Owner: ruoqizzz
• License: mit
• Created: 2019-11-08T18:07:49.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-02-02T10:34:04.000Z (over 4 years ago)
• Last Synced: 2024-05-12T02:30:08.116Z (about 1 month ago)
• Language: Jupyter Notebook
• Size: 27.6 MB
• Stars: 11
• Watchers: 2
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• awesome-stars - Scale-invariant-CNNs - Study the Scale Invariance or Equivariance Convolutional Neural Network (Jupyter Notebook)

README

        
# Scale Invariance/Equavariance Convolutional Neural Network 

This project is focus on evaluating two to three recent approaches to achive scale equivariance and/or invariance of CNNs.

## Paper:

1. [Locally Scale-Invariant Convolutional Neural Network](https://arxiv.org/abs/1412.5104)

   - **Method**: Firstly,  they applies filters at multiple scales in each layer so a single filter can detect and learn patterns at multiple scales. Then, 	max-pool responses over scales to obtain representations that are locally scale invariant yet have the same dimensionality as a traditional ConvNet layer output.

   - **Dataset**: MNIST-Scale

2. [Scale Steerable Filters for Locally Scale-Invariant Convolutional Neural Networks](https://arxiv.org/abs/1906.03861)

   - **Method**: Using the log-radial harmonics as a complex steerable basis, we construct a lo-

     cally scale invariant CNN, where the filters in each convolution layer are a linear combination of the basis filters.

   - **Dataset**: MNIST-Scale

3.  [Making Convolutional Network Shift-Invariant Again](https://arxiv.org/abs/1904.11486)

      - **Method**: Antialiasing filter combined with subsampling, for example, max pooling and CNN with stride. 

      - **Dataset**: MNIST-Scale

 

## Schedule

- **11 Nov - 24 Nov**:

  - Write the summary of *Locally Scale-Invariant Convolutional Neural Network*. 

  - Implement the results of *Locally Scale-Invariant Convolutional Neural Network on MNIST-Scale dataset*.

- **25 Nov - 08 Dec**:

    - Write the summary of *Scale Steerable Filters for Locally Scale-Invariant Convolutional Neural Networks*.

    - Implement the results of *Scale Steerable Filters for Locally Scale-Invariant Convolutional Neural Networks* on MNIST-Scale.

- **09 Dec - 22 Dec**:

    - Write the summary of *Making Convolutional Network Shift-Invatiant Again*

    - Combine the method with SS-CNN, denoted as SS-CNN-BlurPool

    - Evaluate the method on MNIST-Scale.

    - Implement the baseline CNN on MNIST-Scale

    - Compare the results of CNN, SS-CNN, SI-ConvNet, and SS-CNN-BlurPool.

- **23 Dec - -5 Jan**:

   - Preproccessing with dataset Oral Cancer

   - Evaluation on different training size

- **06 Jan - 12 Jan**:

    - Write the report. 

    - Design poster.