libavfilter/dnn: support multiple outputs for tensorflow model

some models such as ssd, yolo have more than one output.

the clean up code in this patch is a little complex, it is because
that set_input_output_tf could be called for many times together
with ff_dnn_execute_model_tf, we have to clean resources for the
case that the two interfaces are called interleaved.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Pedro Arthur <bygrandao@gmail.com>

1 files changed, 9 insertions, 6 deletions

diff --git a/libavfilter/dnn_backend_native.c b/libavfilter/dnn_backend_native.c
index 18735c025c..8a83c63c73 100644
--- a/libavfilter/dnn_backend_native.c
+++ b/libavfilter/dnn_backend_native.c
@@ -25,7 +25,7 @@
 
 #include "dnn_backend_native.h"
 
-static DNNReturnType set_input_output_native(void *model, DNNData *input, const char *input_name, const char *output_name)
+static DNNReturnType set_input_output_native(void *model, DNNData *input, const char *input_name, const char **output_names, uint32_t nb_output)
 {
     ConvolutionalNetwork *network = (ConvolutionalNetwork *)model;
     InputParams *input_params;
@@ -275,7 +275,7 @@ static void depth_to_space(const float *input, float *output, int block_size, in
     }
 }
 
-DNNReturnType ff_dnn_execute_model_native(const DNNModel *model, DNNData *output)
+DNNReturnType ff_dnn_execute_model_native(const DNNModel *model, DNNData *outputs, uint32_t nb_output)
 {
     ConvolutionalNetwork *network = (ConvolutionalNetwork *)model->model;
     int cur_width, cur_height, cur_channels;
@@ -317,10 +317,13 @@ DNNReturnType ff_dnn_execute_model_native(const DNNModel *model, DNNData *output
         }
     }
 
-    output->data = network->layers[network->layers_num - 1].output;
-    output->height = cur_height;
-    output->width = cur_width;
-    output->channels = cur_channels;
+    // native mode does not support multiple outputs yet
+    if (nb_output > 1)
+        return DNN_ERROR;
+    outputs[0].data = network->layers[network->layers_num - 1].output;
+    outputs[0].height = cur_height;
+    outputs[0].width = cur_width;
+    outputs[0].channels = cur_channels;
 
     return DNN_SUCCESS;
 }