History

avfilter/dnn: get the data type of network output from dnn execution result

so, we can make a filter more general to accept different network models, by adding a data type convertion after getting data from network. After we add dt field into struct DNNData, it becomes the same as DNNInputData, so merge them with one struct: DNNData.

Guo, Yejun committed 7 years ago

Commit e1b45b81 parent dff39ea8 files · +13 −13Browse files

Showing the changes to libavfilter (8 of 8 files). Show all

Unified · Split

8 files
  1. libavfilter/dnn/dnn_backend_native.c +2 −1
  2. libavfilter/dnn/dnn_backend_native_layer_conv2d.c +1 −0
  3. libavfilter/dnn/dnn_backend_native_layer_depth2space.c +1 −0
  4. libavfilter/dnn/dnn_backend_native_layer_pad.c +1 −0
  5. libavfilter/dnn/dnn_backend_tf.c +3 −2
  6. libavfilter/dnn_interface.h +2 −7
  7. libavfilter/vf_derain.c +2 −2
  8. libavfilter/vf_sr.c +1 −1
libavfilter/dnn/dnn_backend_native.c+2−1
@@ -28,7 +28,7 @@ · 27 unmodified lines
2828 #include "dnn_backend_native_layer_conv2d.h"
2929 #include "dnn_backend_native_layers.h"
3030
31−static DNNReturnType set_input_output_native(void *model, DNNInputData *input, const char *input_name, const char **output_names, uint32_t nb_output)
31+static DNNReturnType set_input_output_native(void *model, DNNData *input, const char *input_name, const char **output_names, uint32_t nb_output)
3232 {
3333 ConvolutionalNetwork *network = (ConvolutionalNetwork *)model;
3434 DnnOperand *oprd = NULL;
@@ -263,6 +263,7 @@ DNNReturnType ff_dnn_execute_model_native(const DNNModel *model, DNNData *output · 228 unmodified lines
263263 outputs[i].height = oprd->dims[1];
264264 outputs[i].width = oprd->dims[2];
265265 outputs[i].channels = oprd->dims[3];
266+ outputs[i].dt = oprd->data_type;
266267 }
267268
268269 return DNN_SUCCESS;
libavfilter/dnn/dnn_backend_native_layer_conv2d.c+1−0
@@ -106,6 +106,7 @@ int dnn_execute_layer_conv2d(DnnOperand *operands, const int32_t *input_operand_ · 105 unmodified lines
106106 output_operand->dims[1] = height - pad_size * 2;
107107 output_operand->dims[2] = width - pad_size * 2;
108108 output_operand->dims[3] = conv_params->output_num;
109+ output_operand->data_type = operands[input_operand_index].data_type;
109110 output_operand->length = calculate_operand_data_length(output_operand);
110111 output_operand->data = av_realloc(output_operand->data, output_operand->length);
111112 if (!output_operand->data)
libavfilter/dnn/dnn_backend_native_layer_depth2space.c+1−0
@@ -69,6 +69,7 @@ int dnn_execute_layer_depth2space(DnnOperand *operands, const int32_t *input_ope · 68 unmodified lines
6969 output_operand->dims[1] = height * block_size;
7070 output_operand->dims[2] = width * block_size;
7171 output_operand->dims[3] = new_channels;
72+ output_operand->data_type = operands[input_operand_index].data_type;
7273 output_operand->length = calculate_operand_data_length(output_operand);
7374 output_operand->data = av_realloc(output_operand->data, output_operand->length);
7475 if (!output_operand->data)
libavfilter/dnn/dnn_backend_native_layer_pad.c+1−0
@@ -105,6 +105,7 @@ int dnn_execute_layer_pad(DnnOperand *operands, const int32_t *input_operand_ind · 104 unmodified lines
105105 output_operand->dims[1] = new_height;
106106 output_operand->dims[2] = new_width;
107107 output_operand->dims[3] = new_channel;
108+ output_operand->data_type = operands[input_operand_index].data_type;
108109 output_operand->length = calculate_operand_data_length(output_operand);
109110 output_operand->data = av_realloc(output_operand->data, output_operand->length);
110111 if (!output_operand->data)
libavfilter/dnn/dnn_backend_tf.c+3−2
@@ -83,7 +83,7 @@ static TF_Buffer *read_graph(const char *model_filename) · 82 unmodified lines
8383 return graph_buf;
8484 }
8585
86−static TF_Tensor *allocate_input_tensor(const DNNInputData *input)
86+static TF_Tensor *allocate_input_tensor(const DNNData *input)
8787 {
8888 TF_DataType dt;
8989 size_t size;
@@ -105,7 +105,7 @@ static TF_Tensor *allocate_input_tensor(const DNNInputData *input) · 15 unmodified lines
105105 input_dims[1] * input_dims[2] * input_dims[3] * size);
106106 }
107107
108−static DNNReturnType set_input_output_tf(void *model, DNNInputData *input, const char *input_name, const char **output_names, uint32_t nb_output)
108+static DNNReturnType set_input_output_tf(void *model, DNNData *input, const char *input_name, const char **output_names, uint32_t nb_output)
109109 {
110110 TFModel *tf_model = (TFModel *)model;
111111 TF_SessionOptions *sess_opts;
@@ -603,6 +603,7 @@ DNNReturnType ff_dnn_execute_model_tf(const DNNModel *model, DNNData *outputs, u · 491 unmodified lines
603603 outputs[i].width = TF_Dim(tf_model->output_tensors[i], 2);
604604 outputs[i].channels = TF_Dim(tf_model->output_tensors[i], 3);
605605 outputs[i].data = TF_TensorData(tf_model->output_tensors[i]);
606+ outputs[i].dt = TF_TensorType(tf_model->output_tensors[i]);
606607 }
607608
608609 return DNN_SUCCESS;
libavfilter/dnn_interface.h+2−7
@@ -34,15 +34,10 @@ typedef enum {DNN_NATIVE, DNN_TF} DNNBackendType; · 33 unmodified lines
3434
3535 typedef enum {DNN_FLOAT = 1, DNN_UINT8 = 4} DNNDataType;
3636
37−typedef struct DNNInputData{
37+typedef struct DNNData{
3838 void *data;
3939 DNNDataType dt;
4040 int width, height, channels;
41−} DNNInputData;
42−
43−typedef struct DNNData{
44− float *data;
45− int width, height, channels;
4641 } DNNData;
4742
4843 typedef struct DNNModel{
@@ -50,7 +45,7 @@ typedef struct DNNModel{ · 1 unmodified line
5045 void *model;
5146 // Sets model input and output.
5247 // Should be called at least once before model execution.
53− DNNReturnType (*set_input_output)(void *model, DNNInputData *input, const char *input_name, const char **output_names, uint32_t nb_output);
48+ DNNReturnType (*set_input_output)(void *model, DNNData *input, const char *input_name, const char **output_names, uint32_t nb_output);
5449 } DNNModel;
5550
5651 // Stores pointers to functions for loading, executing, freeing DNN models for one of the backends.
libavfilter/vf_derain.c+2−2
@@ -39,7 +39,7 @@ typedef struct DRContext { · 38 unmodified lines
3939 DNNBackendType backend_type;
4040 DNNModule *dnn_module;
4141 DNNModel *model;
42− DNNInputData input;
42+ DNNData input;
4343 DNNData output;
4444 } DRContext;
4545
@@ -137,7 +137,7 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in) · 91 unmodified lines
137137 int t = i * out->width * 3 + j;
138138
139139 int t_in = (i + pad_size) * in->width * 3 + j + pad_size * 3;
140− out->data[0][k] = CLIP((int)((((float *)dr_context->input.data)[t_in] - dr_context->output.data[t]) * 255), 0, 255);
140+ out->data[0][k] = CLIP((int)((((float *)dr_context->input.data)[t_in] - ((float *)dr_context->output.data)[t]) * 255), 0, 255);
141141 }
142142 }
143143
libavfilter/vf_sr.c+1−1
@@ -41,7 +41,7 @@ typedef struct SRContext { · 40 unmodified lines
4141 DNNBackendType backend_type;
4242 DNNModule *dnn_module;
4343 DNNModel *model;
44− DNNInputData input;
44+ DNNData input;
4545 DNNData output;
4646 int scale_factor;
4747 struct SwsContext *sws_contexts[3];

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t