layers3.Layer(*args, **kwargs) |
Layer template. |
layers3.conv1D(size, outchn[, stride, pad, …]) |
Basic convolution 1D layer |
layers3.conv2D(size, outchn[, stride, pad, …]) |
Basic convolution 2D layer |
layers3.conv3D(size, outchn[, stride, pad, …]) |
Basic convolution 3D layer |
layers3.dwconv2D(size, multiplier[, stride, …]) |
Basic depth-wise convolution layer. |
layers3.deconv1D(size, outchn[, stride, …]) |
Basic transposed convolution 1D layer |
layers3.deconv2D(size, outchn[, stride, …]) |
Basic transposed convolution 2D layer |
layers3.deconv3D(size, outchn[, stride, …]) |
Basic transposed convolution 3D layer |
layers3.maxpoolLayer(size, stride[, pad]) |
Basic max pooling layer |
layers3.avgpoolLayer(size, stride[, pad]) |
Basic average pooling layer |
layers3.globalAvgpoolLayer() |
Basic global average pooling layer |
layers3.activation(param, **kwargs) |
Basic activation layer |
layers3.fcLayer(outsize[, usebias, values, norm]) |
Basic fully connected layer |
layers3.batch_norm([decay, epsilon, …]) |
Basic batch normalization layer |
layers3.flatten() |
Basic flatten layer |
layers3.graphConvLayer(outsize[, adj_mtx, …]) |
Basic graph convolution layer |
layers3.bilinearUpSample(factor) |
Basic bilinear upsampling layer. |
layers3.NALU(outdim) |
Arxiv: 1808.00508 |