diff options
author | Guo, Yejun <yejun.guo@intel.com> | 2019-09-05 14:00:28 +0800 |
---|---|---|
committer | Pedro Arthur <bygrandao@gmail.com> | 2019-09-19 11:09:25 -0300 |
commit | 5f058dd693c4bebcd6a293da4630441f3540902f (patch) | |
tree | 4cf12f2fac688758369e55690afb434afc3c947d /libavfilter/dnn/dnn_backend_native.c | |
parent | c2ab998ff38fa11092ccb1c51ab0a1fe9c24ab09 (diff) |
libavfilter/dnn: separate conv2d layer from dnn_backend_native.c to a new file
the logic is that one layer in one separated source file to make
the source files simple for maintaining.
Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>
Diffstat (limited to 'libavfilter/dnn/dnn_backend_native.c')
-rw-r--r-- | libavfilter/dnn/dnn_backend_native.c | 80 |
1 files changed, 1 insertions, 79 deletions
diff --git a/libavfilter/dnn/dnn_backend_native.c b/libavfilter/dnn/dnn_backend_native.c index f56cd81187..5dabd151c0 100644 --- a/libavfilter/dnn/dnn_backend_native.c +++ b/libavfilter/dnn/dnn_backend_native.c @@ -26,6 +26,7 @@ #include "dnn_backend_native.h" #include "libavutil/avassert.h" #include "dnn_backend_native_layer_pad.h" +#include "dnn_backend_native_layer_conv2d.h" static DNNReturnType set_input_output_native(void *model, DNNInputData *input, const char *input_name, const char **output_names, uint32_t nb_output) { @@ -281,85 +282,6 @@ DNNModel *ff_dnn_load_model_native(const char *model_filename) return model; } -#define CLAMP_TO_EDGE(x, w) ((x) < 0 ? 0 : ((x) >= (w) ? (w - 1) : (x))) - -static int convolve(DnnOperand *operands, const int32_t *input_operand_indexes, int32_t output_operand_index, const ConvolutionalParams *conv_params) -{ - float *output; - int32_t input_operand_index = input_operand_indexes[0]; - int number = operands[input_operand_index].dims[0]; - int height = operands[input_operand_index].dims[1]; - int width = operands[input_operand_index].dims[2]; - int channel = operands[input_operand_index].dims[3]; - const float *input = operands[input_operand_index].data; - - int radius = conv_params->kernel_size >> 1; - int src_linesize = width * conv_params->input_num; - int filter_linesize = conv_params->kernel_size * conv_params->input_num; - int filter_size = conv_params->kernel_size * filter_linesize; - int pad_size = (conv_params->padding_method == VALID) ? (conv_params->kernel_size - 1) / 2 * conv_params->dilation : 0; - - DnnOperand *output_operand = &operands[output_operand_index]; - output_operand->dims[0] = number; - output_operand->dims[1] = height - pad_size * 2; - output_operand->dims[2] = width - pad_size * 2; - output_operand->dims[3] = conv_params->output_num; - output_operand->length = calculate_operand_data_length(output_operand); - output_operand->data = av_realloc(output_operand->data, output_operand->length); - if (!output_operand->data) - return -1; - output = output_operand->data; - - av_assert0(channel == conv_params->input_num); - - for (int y = pad_size; y < height - pad_size; ++y) { - for (int x = pad_size; x < width - pad_size; ++x) { - for (int n_filter = 0; n_filter < conv_params->output_num; ++n_filter) { - output[n_filter] = conv_params->biases[n_filter]; - - for (int ch = 0; ch < conv_params->input_num; ++ch) { - for (int kernel_y = 0; kernel_y < conv_params->kernel_size; ++kernel_y) { - for (int kernel_x = 0; kernel_x < conv_params->kernel_size; ++kernel_x) { - float input_pel; - if (conv_params->padding_method == SAME_CLAMP_TO_EDGE) { - int y_pos = CLAMP_TO_EDGE(y + (kernel_y - radius) * conv_params->dilation, height); - int x_pos = CLAMP_TO_EDGE(x + (kernel_x - radius) * conv_params->dilation, width); - input_pel = input[y_pos * src_linesize + x_pos * conv_params->input_num + ch]; - } else { - int y_pos = y + (kernel_y - radius) * conv_params->dilation; - int x_pos = x + (kernel_x - radius) * conv_params->dilation; - input_pel = (x_pos < 0 || x_pos >= width || y_pos < 0 || y_pos >= height) ? 0.0 : - input[y_pos * src_linesize + x_pos * conv_params->input_num + ch]; - } - - - output[n_filter] += input_pel * conv_params->kernel[n_filter * filter_size + kernel_y * filter_linesize + - kernel_x * conv_params->input_num + ch]; - } - } - } - switch (conv_params->activation){ - case RELU: - output[n_filter] = FFMAX(output[n_filter], 0.0); - break; - case TANH: - output[n_filter] = 2.0f / (1.0f + exp(-2.0f * output[n_filter])) - 1.0f; - break; - case SIGMOID: - output[n_filter] = 1.0f / (1.0f + exp(-output[n_filter])); - break; - case NONE: - break; - case LEAKY_RELU: - output[n_filter] = FFMAX(output[n_filter], 0.0) + 0.2 * FFMIN(output[n_filter], 0.0); - } - } - output += conv_params->output_num; - } - } - return 0; -} - static int depth_to_space(DnnOperand *operands, const int32_t *input_operand_indexes, int32_t output_operand_index, int block_size) { float *output; |