Michalis Spyrou | 780db4e | 2017-11-23 09:49:51 +0000 | [diff] [blame^] | 1 | /* |
| 2 | * Copyright (c) 2017, 2018 ARM Limited. |
| 3 | * |
| 4 | * SPDX-License-Identifier: MIT |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to |
| 8 | * deal in the Software without restriction, including without limitation the |
| 9 | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in all |
| 14 | * copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | * SOFTWARE. |
| 23 | */ |
| 24 | #ifndef __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__ |
| 25 | #define __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__ |
| 26 | |
| 27 | #include "arm_compute/runtime/CL/functions/CLDeconvolutionLayerUpsample.h" |
| 28 | #include "arm_compute/runtime/CL/functions/CLDirectConvolutionLayer.h" |
| 29 | |
| 30 | #include "arm_compute/runtime/CL/CLMemoryGroup.h" |
| 31 | #include "arm_compute/runtime/CL/CLTensor.h" |
| 32 | #include "arm_compute/runtime/IFunction.h" |
| 33 | #include "arm_compute/runtime/IMemoryManager.h" |
| 34 | |
| 35 | #include <memory> |
| 36 | |
| 37 | namespace arm_compute |
| 38 | { |
| 39 | class ICLTensor; |
| 40 | /** Function to run the deconvolution layer. |
| 41 | * |
| 42 | * Deconvolution Layer is the backward pass of Convolution Layer. First we transform the input depending on the stride and pad info and then perform a 1x1 |
| 43 | * convolution pass. Input stride defines how many zeroes we should put between each element of the input, pad is the amount of padding and finally a is a user |
| 44 | * specified value where a < stride - 1, that increases the padding top and right of the input image. |
| 45 | * |
| 46 | * The relation between input to output is as follows: |
| 47 | * width_output = round((width_input − 1) ∗ (stride_x - 1) − 2 ∗ padding_x + kernel_x + inner_border_right ) |
| 48 | * height_output = round((height_input − 1) ∗ (stride_y - 1) − 2 ∗ padding_y + kernel_y + inner_border_top ) |
| 49 | * |
| 50 | * where: |
| 51 | * width_input is the size of the first input dimension. |
| 52 | * height_input is the size of the second input dimension. |
| 53 | * width_output is the size of the first output dimension. |
| 54 | * height_output is the size of the second output dimension. |
| 55 | * kernel_x and kernel_y are the convolution sizes in x and y. |
| 56 | * inner_border_right and inner_border_top the number of zeros added to the right and top edges of the input. |
| 57 | * stride_x and stride_y is the input stride of the first and second dimension. |
| 58 | * |
| 59 | */ |
| 60 | class CLDeconvolutionLayer : public IFunction |
| 61 | { |
| 62 | public: |
| 63 | /** Constructor */ |
| 64 | CLDeconvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager = nullptr); |
| 65 | /** Set the input, weights, biases and output tensors. |
| 66 | * |
| 67 | * @param[in,out] input Input tensor. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: F32. |
| 68 | * @param[in] weights The 4d weights with dimensions [width, height, OFM, IFM]. Data type supported: Same as @p input. |
| 69 | * @param[in] bias (Optional) The biases have one dimension. Data type supported: Same as @p input. |
| 70 | * @param[out] output Output tensor. The output has the same number of dimensions as the @p input. |
| 71 | * @param[in] info Contains padding and policies to be used in the deconvolution, this is decribed in @ref PadStrideInfo. |
| 72 | * @param[in] inner_border_right The number of zeros added to right edge of the input. |
| 73 | * @param[in] inner_border_top The number of zeros added to top edge of the input. |
| 74 | * |
| 75 | */ |
| 76 | void configure(ICLTensor *input, const ICLTensor *weights, const ICLTensor *bias, ICLTensor *output, const PadStrideInfo &info, |
| 77 | unsigned int inner_border_right, unsigned int inner_border_top); |
| 78 | /** Static function to check if given info will lead to a valid configuration of @ref CLDeconvolutionLayer |
| 79 | * |
| 80 | * @param[in] input Input tensor info. 3 lower dimensions represent a single input, and an optional 4th dimension for batch of inputs. Data types supported: F32. |
| 81 | * @param[in] weights The 4d weights info with dimensions [width, height, OFM, IFM]. Data type supported: Same as @p input. |
| 82 | * @param[in] bias (Optional) The biases have one dimension. Data type supported: Same as @p input. |
| 83 | * @param[in] output Output tensor info. The output has the same number of dimensions as the @p input. |
| 84 | * @param[in] info Contains padding and policies to be used in the deconvolution, this is decribed in @ref PadStrideInfo. |
| 85 | * @param[in] inner_border_right The number of zeros added to right edge of the input. |
| 86 | * @param[in] inner_border_top The number of zeros added to top edge of the input. |
| 87 | * |
| 88 | * @return a status |
| 89 | */ |
| 90 | static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *output, const PadStrideInfo &info, |
| 91 | unsigned int inner_border_right, unsigned int inner_border_top); |
| 92 | |
| 93 | // Inherited methods overridden: |
| 94 | void run() override; |
| 95 | |
| 96 | private: |
| 97 | CLMemoryGroup _memory_group; |
| 98 | CLDeconvolutionLayerUpsample _scale_f; |
| 99 | CLDirectConvolutionLayer _conv_f; |
| 100 | CLTensor _scaled_output; |
| 101 | }; |
| 102 | } |
| 103 | #endif /* __ARM_COMPUTE_CLDECONVOLUTIONLAYER_H__ */ |