This project is focus on evaluating two to three recent approaches to achive scale equivariance and/or invariance of CNNs.
- Locally Scale-Invariant Convolutional Neural Network
- 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
- Scale Steerable Filters for Locally Scale-Invariant Convolutional Neural Networks
- 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
- Making Convolutional Network Shift-Invariant Again
- Method: Antialiasing filter combined with subsampling, for example, max pooling and CNN with stride.
- Dataset: MNIST-Scale
-
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.