From b9070a42a44ec1a0102e2f0b04523d2e96392903 Mon Sep 17 00:00:00 2001 From: Matthew Jackson Date: Thu, 22 Aug 2019 16:13:27 +0100 Subject: COMPMID-2605: Add asymmetric padding support for Deconvolution layer Change-Id: I63b773bdce25f1342ccd3a08ded623a1508f70fe Signed-off-by: Matthew Jackson Reviewed-on: https://review.mlplatform.org/c/1797 Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins Reviewed-by: Michele Di Giorgio Reviewed-by: Giuseppe Rossini --- .../NEON/functions/NEDeconvolutionLayer.cpp | 72 +++++++++++++++------- 1 file changed, 51 insertions(+), 21 deletions(-) (limited to 'src/runtime/NEON/functions/NEDeconvolutionLayer.cpp') diff --git a/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp b/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp index 1f2cc3d73b..bbb91b4651 100644 --- a/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp +++ b/src/runtime/NEON/functions/NEDeconvolutionLayer.cpp @@ -64,13 +64,8 @@ Status NEDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInf const unsigned int height_idx = get_data_layout_dimension_index(weights->data_layout(), DataLayoutDimension::HEIGHT); ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(width_idx) != weights->dimension(height_idx)); ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(width_idx) < 1); - ARM_COMPUTE_RETURN_ERROR_ON(!info.padding_is_symmetric()); - const unsigned int stride_x = info.stride().first; - const unsigned int stride_y = info.stride().second; - - auto out_dims = deconvolution_output_dimensions(input->dimension(width_idx), input->dimension(height_idx), weights->dimension(width_idx), weights->dimension(height_idx), - info.pad().first, info.pad().second, stride_x, stride_y); + auto out_dims = deconvolution_output_dimensions(input->dimension(width_idx), input->dimension(height_idx), weights->dimension(width_idx), weights->dimension(height_idx), info); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, weights); if(bias != nullptr) @@ -96,9 +91,11 @@ Status NEDeconvolutionLayer::validate(const ITensorInfo *input, const ITensorInf ARM_COMPUTE_RETURN_ERROR_ON_MSG(output->dimension(Window::DimZ) != output_shape.z(), "Output's depth is invalid."); } - unsigned int padx = 0; - unsigned int pady = 0; - const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input, *weights, stride_x, stride_y, out_dims, padx, pady); + unsigned int deconv_pad_x = 0; + unsigned int deconv_pad_y = 0; + const unsigned int stride_x = info.stride().first; + const unsigned int stride_y = info.stride().second; + const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input, *weights, stride_x, stride_y, out_dims, deconv_pad_x, deconv_pad_y); TensorInfo scale_out_info(input->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(scale_out_shape)); const PadStrideInfo conv_info(1, 1, 0, 0, 0, 0, DimensionRoundingType::CEIL); @@ -126,14 +123,17 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con _is_prepared = false; _is_nchw = data_layout == DataLayout::NCHW; + const unsigned int pad_left = info.pad_left(); + const unsigned int pad_right = info.pad_right(); + const unsigned int pad_top = info.pad_top(); + const unsigned int pad_bottom = info.pad_bottom(); const unsigned int stride_x = info.stride().first; const unsigned int stride_y = info.stride().second; const unsigned int width_idx = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH); const unsigned int height_idx = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT); - auto out_dims = deconvolution_output_dimensions(input->info()->dimension(width_idx), input->info()->dimension(height_idx), weights->info()->dimension(width_idx), - weights->info()->dimension(height_idx), - info.pad().first, info.pad().second, stride_x, stride_y); + auto out_dims = deconvolution_output_dimensions(input->info()->dimension(width_idx), input->info()->dimension(height_idx), + weights->info()->dimension(width_idx), weights->info()->dimension(height_idx), info); const TensorShape output_shape = compute_deconvolution_output_shape(out_dims, *input->info(), *weights->info()); // Output auto initialization if not yet initialized @@ -157,16 +157,30 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con _permuted_weights.info()->set_data_layout(DataLayout::NCHW); // Find the upsampled dimensions and the padding needed for the convolution with stride 1 in order to match output shape - unsigned int padx = 0; - unsigned int pady = 0; - const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*_permuted_input.info(), *_permuted_weights.info(), stride_x, stride_y, out_dims, padx, - pady); + unsigned int deconv_pad_x = 0; + unsigned int deconv_pad_y = 0; + const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*_permuted_input.info(), *_permuted_weights.info(), stride_x, stride_y, out_dims, + deconv_pad_x, deconv_pad_y); + + unsigned int deconv_pad_left = pad_right > pad_left ? pad_right - pad_left : 0; + unsigned int deconv_pad_right = pad_left > pad_right ? pad_left - pad_right : 0; + deconv_pad_x -= deconv_pad_left + deconv_pad_right; + ARM_COMPUTE_ERROR_ON((deconv_pad_x % 2) != 0); + deconv_pad_left += deconv_pad_x / 2; + deconv_pad_right += deconv_pad_x / 2; + + unsigned int deconv_pad_top = pad_bottom > pad_top ? pad_bottom - pad_top : 0; + unsigned int deconv_pad_bottom = pad_top > pad_bottom ? pad_top - pad_bottom : 0; + deconv_pad_y -= deconv_pad_top + deconv_pad_bottom; + ARM_COMPUTE_ERROR_ON((deconv_pad_y % 2) != 0); + deconv_pad_top += deconv_pad_y / 2; + deconv_pad_bottom += deconv_pad_y / 2; TensorInfo scale_out_info(scale_out_shape, 1, _permuted_input.info()->data_type(), _permuted_input.info()->quantization_info()); scale_out_info.set_data_layout(DataLayout::NCHW); _scaled_output.allocator()->init(scale_out_info); - const PadStrideInfo upsample_info(stride_x, stride_y, padx / 2, pady / 2); + const PadStrideInfo upsample_info(stride_x, stride_y, deconv_pad_left, deconv_pad_right, deconv_pad_top, deconv_pad_bottom, DimensionRoundingType::FLOOR); _upsample_f.configure(&_permuted_input, &_scaled_output, upsample_info); _weights_flipped.allocator()->init(*_permuted_weights.info()->clone()); @@ -189,14 +203,30 @@ void NEDeconvolutionLayer::configure(ITensor *input, const ITensor *weights, con else { // Find the upsampled dimensions and the padding needed for the convolution with stride 1 in order to match output shape - unsigned int padx = 0; - unsigned int pady = 0; - const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input->info(), *weights->info(), stride_x, stride_y, out_dims, padx, pady); + unsigned int deconv_pad_x = 0; + unsigned int deconv_pad_y = 0; + const TensorShape scale_out_shape = compute_deconvolution_upsampled_shape(*input->info(), *weights->info(), stride_x, stride_y, + out_dims, deconv_pad_x, deconv_pad_y); + + unsigned int deconv_pad_left = pad_right > pad_left ? pad_right - pad_left : 0; + unsigned int deconv_pad_right = pad_left > pad_right ? pad_left - pad_right : 0; + deconv_pad_x -= deconv_pad_left + deconv_pad_right; + ARM_COMPUTE_ERROR_ON((deconv_pad_x % 2) != 0); + deconv_pad_left += deconv_pad_x / 2; + deconv_pad_right += deconv_pad_x / 2; + + unsigned int deconv_pad_top = pad_bottom > pad_top ? pad_bottom - pad_top : 0; + unsigned int deconv_pad_bottom = pad_top > pad_bottom ? pad_top - pad_bottom : 0; + deconv_pad_y -= deconv_pad_top + deconv_pad_bottom; + ARM_COMPUTE_ERROR_ON((deconv_pad_y % 2) != 0); + deconv_pad_top += deconv_pad_y / 2; + deconv_pad_bottom += deconv_pad_y / 2; TensorInfo scale_out_info(scale_out_shape, 1, input->info()->data_type(), input->info()->quantization_info()); scale_out_info.set_data_layout(data_layout); _scaled_output.allocator()->init(scale_out_info); - const PadStrideInfo upsample_info(stride_x, stride_y, padx / 2, pady / 2); + + const PadStrideInfo upsample_info(stride_x, stride_y, deconv_pad_left, deconv_pad_right, deconv_pad_top, deconv_pad_bottom, DimensionRoundingType::FLOOR); _upsample_f.configure(input, &_scaled_output, upsample_info); _weights_flipped.allocator()->init(weights->info()->clone()->set_data_layout(data_layout)); -- cgit v1.2.1