dnn: change dnn interface to replace DNNData* with AVFrame*

Currently, every filter needs to provide code to transfer data from
AVFrame* to model input (DNNData*), and also from model output
(DNNData*) to AVFrame*. Actually, such transfer can be implemented
within DNN module, and so filter can focus on its own business logic.

DNN module also exports the function pointer pre_proc and post_proc
in struct DNNModel, just in case that a filter has its special logic
to transfer data between AVFrame* and DNNData*. The default implementation
within DNN module is used if the filter does not set pre/post_proc.

1 files changed, 63 insertions, 103 deletions

diff --git a/libavfilter/vf_sr.c b/libavfilter/vf_sr.c
index 445777f0c6..2eda8c3219 100644
--- a/libavfilter/vf_sr.c
+++ b/libavfilter/vf_sr.c
@@ -41,11 +41,10 @@ typedef struct SRContext {
     DNNBackendType backend_type;
     DNNModule *dnn_module;
     DNNModel *model;
-    DNNData input;
-    DNNData output;
     int scale_factor;
-    struct SwsContext *sws_contexts[3];
-    int sws_slice_h, sws_input_linesize, sws_output_linesize;
+    struct SwsContext *sws_uv_scale;
+    int sws_uv_height;
+    struct SwsContext *sws_pre_scale;
 } SRContext;
 
 #define OFFSET(x) offsetof(SRContext, x)
@@ -87,11 +86,6 @@ static av_cold int init(AVFilterContext *context)
         return AVERROR(EIO);
     }
 
-    sr_context->input.dt = DNN_FLOAT;
-    sr_context->sws_contexts[0] = NULL;
-    sr_context->sws_contexts[1] = NULL;
-    sr_context->sws_contexts[2] = NULL;
-
     return 0;
 }
 
@@ -111,95 +105,63 @@ static int query_formats(AVFilterContext *context)
     return ff_set_common_formats(context, formats_list);
 }
 
-static int config_props(AVFilterLink *inlink)
+static int config_output(AVFilterLink *outlink)
 {
-    AVFilterContext *context = inlink->dst;
-    SRContext *sr_context = context->priv;
-    AVFilterLink *outlink = context->outputs[0];
+    AVFilterContext *context = outlink->src;
+    SRContext *ctx = context->priv;
     DNNReturnType result;
-    int sws_src_h, sws_src_w, sws_dst_h, sws_dst_w;
+    AVFilterLink *inlink = context->inputs[0];
+    AVFrame *out = NULL;
     const char *model_output_name = "y";
 
-    sr_context->input.width = inlink->w * sr_context->scale_factor;
-    sr_context->input.height = inlink->h * sr_context->scale_factor;
-    sr_context->input.channels = 1;
-
-    result = (sr_context->model->set_input)(sr_context->model->model, &sr_context->input, "x");
-    if (result != DNN_SUCCESS){
-        av_log(context, AV_LOG_ERROR, "could not set input and output for the model\n");
+    AVFrame *fake_in = ff_get_video_buffer(inlink, inlink->w, inlink->h);
+    result = (ctx->model->set_input)(ctx->model->model, fake_in, "x");
+    if (result != DNN_SUCCESS) {
+        av_log(context, AV_LOG_ERROR, "could not set input for the model\n");
         return AVERROR(EIO);
     }
 
-    result = (sr_context->dnn_module->execute_model)(sr_context->model, &sr_context->output, &model_output_name, 1);
+    // have a try run in case that the dnn model resize the frame
+    out = ff_get_video_buffer(inlink, inlink->w, inlink->h);
+    result = (ctx->dnn_module->execute_model)(ctx->model, (const char **)&model_output_name, 1, out);
     if (result != DNN_SUCCESS){
         av_log(context, AV_LOG_ERROR, "failed to execute loaded model\n");
         return AVERROR(EIO);
     }
 
-    if (sr_context->input.height != sr_context->output.height || sr_context->input.width != sr_context->output.width){
-        sr_context->input.width = inlink->w;
-        sr_context->input.height = inlink->h;
-        result = (sr_context->model->set_input)(sr_context->model->model, &sr_context->input, "x");
-        if (result != DNN_SUCCESS){
-            av_log(context, AV_LOG_ERROR, "could not set input and output for the model\n");
-            return AVERROR(EIO);
-        }
-        result = (sr_context->dnn_module->execute_model)(sr_context->model, &sr_context->output, &model_output_name, 1);
-        if (result != DNN_SUCCESS){
-            av_log(context, AV_LOG_ERROR, "failed to execute loaded model\n");
-            return AVERROR(EIO);
-        }
-        sr_context->scale_factor = 0;
-    }
-    outlink->h = sr_context->output.height;
-    outlink->w = sr_context->output.width;
-    sr_context->sws_contexts[1] = sws_getContext(sr_context->input.width, sr_context->input.height, AV_PIX_FMT_GRAY8,
-                                                 sr_context->input.width, sr_context->input.height, AV_PIX_FMT_GRAYF32,
-                                                 0, NULL, NULL, NULL);
-    sr_context->sws_input_linesize = sr_context->input.width << 2;
-    sr_context->sws_contexts[2] = sws_getContext(sr_context->output.width, sr_context->output.height, AV_PIX_FMT_GRAYF32,
-                                                 sr_context->output.width, sr_context->output.height, AV_PIX_FMT_GRAY8,
-                                                 0, NULL, NULL, NULL);
-    sr_context->sws_output_linesize = sr_context->output.width << 2;
-    if (!sr_context->sws_contexts[1] || !sr_context->sws_contexts[2]){
-        av_log(context, AV_LOG_ERROR, "could not create SwsContext for conversions\n");
-        return AVERROR(ENOMEM);
-    }
-    if (sr_context->scale_factor){
-        sr_context->sws_contexts[0] = sws_getContext(inlink->w, inlink->h, inlink->format,
-                                                     outlink->w, outlink->h, outlink->format,
-                                                     SWS_BICUBIC, NULL, NULL, NULL);
-        if (!sr_context->sws_contexts[0]){
-            av_log(context, AV_LOG_ERROR, "could not create SwsContext for scaling\n");
-            return AVERROR(ENOMEM);
-        }
-        sr_context->sws_slice_h = inlink->h;
-    } else {
+    if (fake_in->width != out->width || fake_in->height != out->height) {
+        //espcn
+        outlink->w = out->width;
+        outlink->h = out->height;
         if (inlink->format != AV_PIX_FMT_GRAY8){
             const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
-            sws_src_h = AV_CEIL_RSHIFT(sr_context->input.height, desc->log2_chroma_h);
-            sws_src_w = AV_CEIL_RSHIFT(sr_context->input.width, desc->log2_chroma_w);
-            sws_dst_h = AV_CEIL_RSHIFT(sr_context->output.height, desc->log2_chroma_h);
-            sws_dst_w = AV_CEIL_RSHIFT(sr_context->output.width, desc->log2_chroma_w);
-
-            sr_context->sws_contexts[0] = sws_getContext(sws_src_w, sws_src_h, AV_PIX_FMT_GRAY8,
-                                                         sws_dst_w, sws_dst_h, AV_PIX_FMT_GRAY8,
-                                                         SWS_BICUBIC, NULL, NULL, NULL);
-            if (!sr_context->sws_contexts[0]){
-                av_log(context, AV_LOG_ERROR, "could not create SwsContext for scaling\n");
-                return AVERROR(ENOMEM);
-            }
-            sr_context->sws_slice_h = sws_src_h;
+            int sws_src_h = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
+            int sws_src_w = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
+            int sws_dst_h = AV_CEIL_RSHIFT(outlink->h, desc->log2_chroma_h);
+            int sws_dst_w = AV_CEIL_RSHIFT(outlink->w, desc->log2_chroma_w);
+            ctx->sws_uv_scale = sws_getContext(sws_src_w, sws_src_h, AV_PIX_FMT_GRAY8,
+                                               sws_dst_w, sws_dst_h, AV_PIX_FMT_GRAY8,
+                                               SWS_BICUBIC, NULL, NULL, NULL);
+            ctx->sws_uv_height = sws_src_h;
         }
+    } else {
+        //srcnn
+        outlink->w = out->width * ctx->scale_factor;
+        outlink->h = out->height * ctx->scale_factor;
+        ctx->sws_pre_scale = sws_getContext(inlink->w, inlink->h, inlink->format,
+                                        outlink->w, outlink->h, outlink->format,
+                                        SWS_BICUBIC, NULL, NULL, NULL);
     }
 
+    av_frame_free(&fake_in);
+    av_frame_free(&out);
     return 0;
 }
 
 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
 {
     AVFilterContext *context = inlink->dst;
-    SRContext *sr_context = context->priv;
+    SRContext *ctx = context->priv;
     AVFilterLink *outlink = context->outputs[0];
     AVFrame *out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
     DNNReturnType dnn_result;
@@ -211,45 +173,44 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
         return AVERROR(ENOMEM);
     }
     av_frame_copy_props(out, in);
-    out->height = sr_context->output.height;
-    out->width = sr_context->output.width;
-    if (sr_context->scale_factor){
-        sws_scale(sr_context->sws_contexts[0], (const uint8_t **)in->data, in->linesize,
-                  0, sr_context->sws_slice_h, out->data, out->linesize);
 
-        sws_scale(sr_context->sws_contexts[1], (const uint8_t **)out->data, out->linesize,
-                  0, out->height, (uint8_t * const*)(&sr_context->input.data),
-                  (const int [4]){sr_context->sws_input_linesize, 0, 0, 0});
+    if (ctx->sws_pre_scale) {
+        sws_scale(ctx->sws_pre_scale,
+                    (const uint8_t **)in->data, in->linesize, 0, in->height,
+                    out->data, out->linesize);
+        dnn_result = (ctx->model->set_input)(ctx->model->model, out, "x");
     } else {
-        if (sr_context->sws_contexts[0]){
-            sws_scale(sr_context->sws_contexts[0], (const uint8_t **)(in->data + 1), in->linesize + 1,
-                      0, sr_context->sws_slice_h, out->data + 1, out->linesize + 1);
-            sws_scale(sr_context->sws_contexts[0], (const uint8_t **)(in->data + 2), in->linesize + 2,
-                      0, sr_context->sws_slice_h, out->data + 2, out->linesize + 2);
-        }
+        dnn_result = (ctx->model->set_input)(ctx->model->model, in, "x");
+    }
 
-        sws_scale(sr_context->sws_contexts[1], (const uint8_t **)in->data, in->linesize,
-                  0, in->height, (uint8_t * const*)(&sr_context->input.data),
-                  (const int [4]){sr_context->sws_input_linesize, 0, 0, 0});
+    if (dnn_result != DNN_SUCCESS) {
+        av_frame_free(&in);
+        av_frame_free(&out);
+        av_log(context, AV_LOG_ERROR, "could not set input for the model\n");
+        return AVERROR(EIO);
     }
-    av_frame_free(&in);
 
-    dnn_result = (sr_context->dnn_module->execute_model)(sr_context->model, &sr_context->output, &model_output_name, 1);
+    dnn_result = (ctx->dnn_module->execute_model)(ctx->model, (const char **)&model_output_name, 1, out);
     if (dnn_result != DNN_SUCCESS){
-        av_log(context, AV_LOG_ERROR, "failed to execute loaded model\n");
+        av_log(ctx, AV_LOG_ERROR, "failed to execute loaded model\n");
+        av_frame_free(&in);
+        av_frame_free(&out);
         return AVERROR(EIO);
     }
 
-    sws_scale(sr_context->sws_contexts[2], (const uint8_t *[4]){(const uint8_t *)sr_context->output.data, 0, 0, 0},
-              (const int[4]){sr_context->sws_output_linesize, 0, 0, 0},
-              0, out->height, (uint8_t * const*)out->data, out->linesize);
+    if (ctx->sws_uv_scale) {
+        sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
+                  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
+        sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 2), in->linesize + 2,
+                  0, ctx->sws_uv_height, out->data + 2, out->linesize + 2);
+    }
 
+    av_frame_free(&in);
     return ff_filter_frame(outlink, out);
 }
 
 static av_cold void uninit(AVFilterContext *context)
 {
-    int i;
     SRContext *sr_context = context->priv;
 
     if (sr_context->dnn_module){
@@ -257,16 +218,14 @@ static av_cold void uninit(AVFilterContext *context)
         av_freep(&sr_context->dnn_module);
     }
 
-    for (i = 0; i < 3; ++i){
-        sws_freeContext(sr_context->sws_contexts[i]);
-    }
+    sws_freeContext(sr_context->sws_uv_scale);
+    sws_freeContext(sr_context->sws_pre_scale);
 }
 
 static const AVFilterPad sr_inputs[] = {
     {
         .name         = "default",
         .type         = AVMEDIA_TYPE_VIDEO,
-        .config_props = config_props,
         .filter_frame = filter_frame,
     },
     { NULL }
@@ -275,6 +234,7 @@ static const AVFilterPad sr_inputs[] = {
 static const AVFilterPad sr_outputs[] = {
     {
         .name = "default",
+        .config_props = config_output,
         .type = AVMEDIA_TYPE_VIDEO,
     },
     { NULL }