FFmpeg
vf_dnn_processing.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2019 Guo Yejun
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 /**
22  * @file
23  * implementing a generic image processing filter using deep learning networks.
24  */
25 
26 #include "libavformat/avio.h"
27 #include "libavutil/opt.h"
28 #include "libavutil/pixdesc.h"
29 #include "libavutil/avassert.h"
30 #include "libavutil/imgutils.h"
31 #include "filters.h"
32 #include "dnn_filter_common.h"
33 #include "formats.h"
34 #include "internal.h"
35 #include "libswscale/swscale.h"
36 #include "libavutil/time.h"
37 
38 typedef struct DnnProcessingContext {
39  const AVClass *class;
44 
45 #define OFFSET(x) offsetof(DnnProcessingContext, dnnctx.x)
46 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM
47 static const AVOption dnn_processing_options[] = {
48  { "dnn_backend", "DNN backend", OFFSET(backend_type), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FLAGS, "backend" },
49  { "native", "native backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FLAGS, "backend" },
50 #if (CONFIG_LIBTENSORFLOW == 1)
51  { "tensorflow", "tensorflow backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, FLAGS, "backend" },
52 #endif
53 #if (CONFIG_LIBOPENVINO == 1)
54  { "openvino", "openvino backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = 2 }, 0, 0, FLAGS, "backend" },
55 #endif
57  { NULL }
58 };
59 
60 AVFILTER_DEFINE_CLASS(dnn_processing);
61 
63 {
65  return ff_dnn_init(&ctx->dnnctx, DFT_PROCESS_FRAME, context);
66 }
67 
69 {
70  static const enum AVPixelFormat pix_fmts[] = {
77  };
79  return ff_set_common_formats(context, fmts_list);
80 }
81 
82 #define LOG_FORMAT_CHANNEL_MISMATCH() \
83  av_log(ctx, AV_LOG_ERROR, \
84  "the frame's format %s does not match " \
85  "the model input channel %d\n", \
86  av_get_pix_fmt_name(fmt), \
87  model_input->channels);
88 
89 static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
90 {
91  AVFilterContext *ctx = inlink->dst;
92  enum AVPixelFormat fmt = inlink->format;
93 
94  // the design is to add explicit scale filter before this filter
95  if (model_input->height != -1 && model_input->height != inlink->h) {
96  av_log(ctx, AV_LOG_ERROR, "the model requires frame height %d but got %d\n",
97  model_input->height, inlink->h);
98  return AVERROR(EIO);
99  }
100  if (model_input->width != -1 && model_input->width != inlink->w) {
101  av_log(ctx, AV_LOG_ERROR, "the model requires frame width %d but got %d\n",
102  model_input->width, inlink->w);
103  return AVERROR(EIO);
104  }
105  if (model_input->dt != DNN_FLOAT) {
106  avpriv_report_missing_feature(ctx, "data type rather than DNN_FLOAT");
107  return AVERROR(EIO);
108  }
109 
110  switch (fmt) {
111  case AV_PIX_FMT_RGB24:
112  case AV_PIX_FMT_BGR24:
113  if (model_input->channels != 3) {
115  return AVERROR(EIO);
116  }
117  return 0;
118  case AV_PIX_FMT_GRAYF32:
119  case AV_PIX_FMT_YUV420P:
120  case AV_PIX_FMT_YUV422P:
121  case AV_PIX_FMT_YUV444P:
122  case AV_PIX_FMT_YUV410P:
123  case AV_PIX_FMT_YUV411P:
124  case AV_PIX_FMT_NV12:
125  if (model_input->channels != 1) {
127  return AVERROR(EIO);
128  }
129  return 0;
130  default:
132  return AVERROR(EIO);
133  }
134 
135  return 0;
136 }
137 
139 {
143  DNNData model_input;
144  int check;
145 
146  result = ff_dnn_get_input(&ctx->dnnctx, &model_input);
147  if (result != DNN_SUCCESS) {
148  av_log(ctx, AV_LOG_ERROR, "could not get input from the model\n");
149  return AVERROR(EIO);
150  }
151 
152  check = check_modelinput_inlink(&model_input, inlink);
153  if (check != 0) {
154  return check;
155  }
156 
157  return 0;
158 }
159 
161 {
163  av_assert0(desc);
164  return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components == 3;
165 }
166 
167 static int prepare_uv_scale(AVFilterLink *outlink)
168 {
169  AVFilterContext *context = outlink->src;
171  AVFilterLink *inlink = context->inputs[0];
172  enum AVPixelFormat fmt = inlink->format;
173 
174  if (isPlanarYUV(fmt)) {
175  if (inlink->w != outlink->w || inlink->h != outlink->h) {
176  if (fmt == AV_PIX_FMT_NV12) {
177  ctx->sws_uv_scale = sws_getContext(inlink->w >> 1, inlink->h >> 1, AV_PIX_FMT_YA8,
178  outlink->w >> 1, outlink->h >> 1, AV_PIX_FMT_YA8,
180  ctx->sws_uv_height = inlink->h >> 1;
181  } else {
183  int sws_src_h = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
184  int sws_src_w = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
185  int sws_dst_h = AV_CEIL_RSHIFT(outlink->h, desc->log2_chroma_h);
186  int sws_dst_w = AV_CEIL_RSHIFT(outlink->w, desc->log2_chroma_w);
187  ctx->sws_uv_scale = sws_getContext(sws_src_w, sws_src_h, AV_PIX_FMT_GRAY8,
188  sws_dst_w, sws_dst_h, AV_PIX_FMT_GRAY8,
190  ctx->sws_uv_height = sws_src_h;
191  }
192  }
193  }
194 
195  return 0;
196 }
197 
198 static int config_output(AVFilterLink *outlink)
199 {
200  AVFilterContext *context = outlink->src;
203  AVFilterLink *inlink = context->inputs[0];
204 
205  // have a try run in case that the dnn model resize the frame
206  result = ff_dnn_get_output(&ctx->dnnctx, inlink->w, inlink->h, &outlink->w, &outlink->h);
207  if (result != DNN_SUCCESS) {
208  av_log(ctx, AV_LOG_ERROR, "could not get output from the model\n");
209  return AVERROR(EIO);
210  }
211 
212  prepare_uv_scale(outlink);
213 
214  return 0;
215 }
216 
218 {
219  const AVPixFmtDescriptor *desc;
220  int uv_height;
221 
222  if (!ctx->sws_uv_scale) {
223  av_assert0(in->height == out->height && in->width == out->width);
225  uv_height = AV_CEIL_RSHIFT(in->height, desc->log2_chroma_h);
226  for (int i = 1; i < 3; ++i) {
227  int bytewidth = av_image_get_linesize(in->format, in->width, i);
228  if (bytewidth < 0) {
229  return AVERROR(EINVAL);
230  }
231  av_image_copy_plane(out->data[i], out->linesize[i],
232  in->data[i], in->linesize[i],
233  bytewidth, uv_height);
234  }
235  } else if (in->format == AV_PIX_FMT_NV12) {
236  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
237  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
238  } else {
239  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
240  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
241  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 2), in->linesize + 2,
242  0, ctx->sws_uv_height, out->data + 2, out->linesize + 2);
243  }
244 
245  return 0;
246 }
247 
249 {
251  AVFilterLink *outlink = context->outputs[0];
253  DNNReturnType dnn_result;
254  AVFrame *out;
255 
256  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
257  if (!out) {
258  av_frame_free(&in);
259  return AVERROR(ENOMEM);
260  }
262 
263  dnn_result = ff_dnn_execute_model(&ctx->dnnctx, in, out);
264  if (dnn_result != DNN_SUCCESS){
265  av_log(ctx, AV_LOG_ERROR, "failed to execute model\n");
266  av_frame_free(&in);
267  av_frame_free(&out);
268  return AVERROR(EIO);
269  }
270 
271  if (isPlanarYUV(in->format))
272  copy_uv_planes(ctx, out, in);
273 
274  av_frame_free(&in);
275  return ff_filter_frame(outlink, out);
276 }
277 
279 {
280  AVFilterLink *inlink = filter_ctx->inputs[0];
281  AVFilterLink *outlink = filter_ctx->outputs[0];
282  AVFrame *in = NULL;
283  int64_t pts;
284  int ret, status;
285  int got_frame = 0;
286 
288 
289  do {
290  // drain all input frames
292  if (ret < 0)
293  return ret;
294  if (ret > 0) {
295  ret = filter_frame(inlink, in);
296  if (ret < 0)
297  return ret;
298  got_frame = 1;
299  }
300  } while (ret > 0);
301 
302  // if frame got, schedule to next filter
303  if (got_frame)
304  return 0;
305 
307  if (status == AVERROR_EOF) {
308  ff_outlink_set_status(outlink, status, pts);
309  return ret;
310  }
311  }
312 
314 
315  return FFERROR_NOT_READY;
316 }
317 
318 static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
319 {
320  DnnProcessingContext *ctx = outlink->src->priv;
321  int ret;
322  DNNAsyncStatusType async_state;
323 
324  ret = ff_dnn_flush(&ctx->dnnctx);
325  if (ret != DNN_SUCCESS) {
326  return -1;
327  }
328 
329  do {
330  AVFrame *in_frame = NULL;
331  AVFrame *out_frame = NULL;
332  async_state = ff_dnn_get_async_result(&ctx->dnnctx, &in_frame, &out_frame);
333  if (out_frame) {
334  if (isPlanarYUV(in_frame->format))
335  copy_uv_planes(ctx, out_frame, in_frame);
336  av_frame_free(&in_frame);
337  ret = ff_filter_frame(outlink, out_frame);
338  if (ret < 0)
339  return ret;
340  if (out_pts)
341  *out_pts = out_frame->pts + pts;
342  }
343  av_usleep(5000);
344  } while (async_state >= DAST_NOT_READY);
345 
346  return 0;
347 }
348 
350 {
351  AVFilterLink *inlink = filter_ctx->inputs[0];
352  AVFilterLink *outlink = filter_ctx->outputs[0];
354  AVFrame *in = NULL, *out = NULL;
355  int64_t pts;
356  int ret, status;
357  int got_frame = 0;
358  int async_state;
359 
361 
362  do {
363  // drain all input frames
365  if (ret < 0)
366  return ret;
367  if (ret > 0) {
368  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
369  if (!out) {
370  av_frame_free(&in);
371  return AVERROR(ENOMEM);
372  }
374  if (ff_dnn_execute_model_async(&ctx->dnnctx, in, out) != DNN_SUCCESS) {
375  return AVERROR(EIO);
376  }
377  }
378  } while (ret > 0);
379 
380  // drain all processed frames
381  do {
382  AVFrame *in_frame = NULL;
383  AVFrame *out_frame = NULL;
384  async_state = ff_dnn_get_async_result(&ctx->dnnctx, &in_frame, &out_frame);
385  if (out_frame) {
386  if (isPlanarYUV(in_frame->format))
387  copy_uv_planes(ctx, out_frame, in_frame);
388  av_frame_free(&in_frame);
389  ret = ff_filter_frame(outlink, out_frame);
390  if (ret < 0)
391  return ret;
392  got_frame = 1;
393  }
394  } while (async_state == DAST_SUCCESS);
395 
396  // if frame got, schedule to next filter
397  if (got_frame)
398  return 0;
399 
401  if (status == AVERROR_EOF) {
402  int64_t out_pts = pts;
403  ret = flush_frame(outlink, pts, &out_pts);
404  ff_outlink_set_status(outlink, status, out_pts);
405  return ret;
406  }
407  }
408 
410 
411  return 0;
412 }
413 
415 {
417 
418  if (ctx->dnnctx.async)
419  return activate_async(filter_ctx);
420  else
421  return activate_sync(filter_ctx);
422 }
423 
425 {
427 
428  sws_freeContext(context->sws_uv_scale);
429  ff_dnn_uninit(&context->dnnctx);
430 }
431 
433  {
434  .name = "default",
435  .type = AVMEDIA_TYPE_VIDEO,
436  .config_props = config_input,
437  },
438  { NULL }
439 };
440 
442  {
443  .name = "default",
444  .type = AVMEDIA_TYPE_VIDEO,
445  .config_props = config_output,
446  },
447  { NULL }
448 };
449 
451  .name = "dnn_processing",
452  .description = NULL_IF_CONFIG_SMALL("Apply DNN processing filter to the input."),
453  .priv_size = sizeof(DnnProcessingContext),
454  .init = init,
455  .uninit = uninit,
459  .priv_class = &dnn_processing_class,
460  .activate = activate,
461 };
ff_get_video_buffer
AVFrame * ff_get_video_buffer(AVFilterLink *link, int w, int h)
Request a picture buffer with a specific set of permissions.
Definition: video.c:99
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_dnn_processing.c:424
dnn_processing_options
static const AVOption dnn_processing_options[]
Definition: vf_dnn_processing.c:47
AVPixelFormat
AVPixelFormat
Pixel format.
Definition: pixfmt.h:64
status
they must not be accessed directly The fifo field contains the frames that are queued in the input for processing by the filter The status_in and status_out fields contains the queued status(EOF or error) of the link
AVERROR
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
opt.h
AV_PIX_FMT_YA8
@ AV_PIX_FMT_YA8
8 bits gray, 8 bits alpha
Definition: pixfmt.h:133
filter_ctx
static FilteringContext * filter_ctx
Definition: transcoding.c:48
ff_make_format_list
AVFilterFormats * ff_make_format_list(const int *fmts)
Create a list of supported formats.
Definition: formats.c:286
ff_dnn_get_output
DNNReturnType ff_dnn_get_output(DnnContext *ctx, int input_width, int input_height, int *output_width, int *output_height)
Definition: dnn_filter_common.c:126
out
FILE * out
Definition: movenc.c:54
OFFSET
#define OFFSET(x)
Definition: vf_dnn_processing.c:45
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:978
av_pix_fmt_desc_get
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:2541
AVERROR_EOF
#define AVERROR_EOF
End of file.
Definition: error.h:55
FFERROR_NOT_READY
return FFERROR_NOT_READY
Definition: filter_design.txt:204
inlink
The exact code depends on how similar the blocks are and how related they are to the and needs to apply these operations to the correct inlink or outlink if there are several Macros are available to factor that when no extra processing is inlink
Definition: filter_design.txt:212
av_frame_free
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:111
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:303
pixdesc.h
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:396
DnnProcessingContext
Definition: vf_dnn_processing.c:38
AVFrame::width
int width
Definition: frame.h:361
AVOption
AVOption.
Definition: opt.h:248
AV_PIX_FMT_BGR24
@ AV_PIX_FMT_BGR24
packed RGB 8:8:8, 24bpp, BGRBGR...
Definition: pixfmt.h:69
ff_dnn_get_input
DNNReturnType ff_dnn_get_input(DnnContext *ctx, DNNData *input)
Definition: dnn_filter_common.c:121
sws_scale
int attribute_align_arg sws_scale(struct SwsContext *c, const uint8_t *const srcSlice[], const int srcStride[], int srcSliceY, int srcSliceH, uint8_t *const dst[], const int dstStride[])
swscale wrapper, so we don't need to export the SwsContext.
Definition: swscale.c:745
DNNData::height
int height
Definition: dnn_interface.h:61
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_dnn_processing.c:138
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:149
dnn_filter_common.h
FF_FILTER_FORWARD_STATUS_BACK
#define FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink)
Forward the status on an output link to an input link.
Definition: filters.h:199
AVFrame::data
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:317
av_image_copy_plane
void av_image_copy_plane(uint8_t *dst, int dst_linesize, const uint8_t *src, int src_linesize, int bytewidth, int height)
Copy image plane from src to dst.
Definition: imgutils.c:374
AVFilterFormats
A list of supported formats for one end of a filter link.
Definition: formats.h:65
formats.h
ff_inlink_consume_frame
int ff_inlink_consume_frame(AVFilterLink *link, AVFrame **rframe)
Take a frame from the link's FIFO and update the link's stats.
Definition: avfilter.c:1376
AVFilterContext::priv
void * priv
private data for use by the filter
Definition: avfilter.h:348
DnnContext
Definition: dnn_filter_common.h:29
DNN_SUCCESS
@ DNN_SUCCESS
Definition: dnn_interface.h:33
pts
static int64_t pts
Definition: transcode_aac.c:652
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:54
flush_frame
static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
Definition: vf_dnn_processing.c:318
avassert.h
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:181
av_cold
#define av_cold
Definition: attributes.h:90
inputs
these buffered frames must be flushed immediately if a new input produces new the filter must not call request_frame to get more It must just process the frame or queue it The task of requesting more frames is left to the filter s request_frame method or the application If a filter has several inputs
Definition: filter_design.txt:243
ff_set_common_formats
int ff_set_common_formats(AVFilterContext *ctx, AVFilterFormats *formats)
A helper for query_formats() which sets all links to the same list of formats.
Definition: formats.c:580
check
#define check(x, y, S, v)
Definition: motion_est_template.c:405
ff_outlink_set_status
static void ff_outlink_set_status(AVFilterLink *link, int status, int64_t pts)
Set the status field of a link from the source filter.
Definition: filters.h:189
copy_uv_planes
static int copy_uv_planes(DnnProcessingContext *ctx, AVFrame *out, const AVFrame *in)
Definition: vf_dnn_processing.c:217
AV_CEIL_RSHIFT
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:58
av_assert0
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
DNNReturnType
DNNReturnType
Definition: dnn_interface.h:33
DNNData
Definition: dnn_interface.h:59
outputs
static const AVFilterPad outputs[]
Definition: af_acontrast.c:203
filters.h
pix_fmts
static enum AVPixelFormat pix_fmts[]
Definition: libkvazaar.c:290
ctx
AVFormatContext * ctx
Definition: movenc.c:48
pix_fmt
static enum AVPixelFormat pix_fmt
Definition: demuxing_decoding.c:41
AV_PIX_FMT_YUV420P
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
Definition: pixfmt.h:66
av_usleep
int av_usleep(unsigned usec)
Sleep for a period of time.
Definition: time.c:84
AV_PIX_FMT_GRAYF32
#define AV_PIX_FMT_GRAYF32
Definition: pixfmt.h:421
check_modelinput_inlink
static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
Definition: vf_dnn_processing.c:89
context
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default minimum maximum flags name is the option keep it simple and lowercase description are in without and describe what they for example set the foo of the bar offset is the offset of the field in your context
Definition: writing_filters.txt:91
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(dnn_processing)
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:67
result
and forward the result(frame or status change) to the corresponding input. If nothing is possible
NULL
#define NULL
Definition: coverity.c:32
av_frame_copy_props
int av_frame_copy_props(AVFrame *dst, const AVFrame *src)
Copy only "metadata" fields from src to dst.
Definition: frame.c:536
init
static av_cold int init(AVFilterContext *context)
Definition: vf_dnn_processing.c:62
time.h
dnn_processing_inputs
static const AVFilterPad dnn_processing_inputs[]
Definition: vf_dnn_processing.c:432
AV_PIX_FMT_GRAY8
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Definition: pixfmt.h:74
ff_inlink_acknowledge_status
int ff_inlink_acknowledge_status(AVFilterLink *link, int *rstatus, int64_t *rpts)
Test and acknowledge the change of status on the link.
Definition: avfilter.c:1331
query_formats
static int query_formats(AVFilterContext *context)
Definition: vf_dnn_processing.c:68
DNN_COMMON_OPTIONS
#define DNN_COMMON_OPTIONS
Definition: dnn_filter_common.h:43
AV_PIX_FMT_RGB24
@ AV_PIX_FMT_RGB24
packed RGB 8:8:8, 24bpp, RGBRGB...
Definition: pixfmt.h:68
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:117
DAST_SUCCESS
@ DAST_SUCCESS
Definition: dnn_interface.h:49
AV_PIX_FMT_FLAG_RGB
#define AV_PIX_FMT_FLAG_RGB
The pixel format contains RGB-like data (as opposed to YUV/grayscale).
Definition: pixdesc.h:137
sws_getContext
struct SwsContext * sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat, int dstW, int dstH, enum AVPixelFormat dstFormat, int flags, SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
Allocate and return an SwsContext.
Definition: utils.c:1910
avio.h
avpriv_report_missing_feature
void avpriv_report_missing_feature(void *avc, const char *msg,...) av_printf_format(2
Log a generic warning message about a missing feature.
AVFrame::format
int format
format of the frame, -1 if unknown or unset Values correspond to enum AVPixelFormat for video frames,...
Definition: frame.h:376
DNNData::dt
DNNDataType dt
Definition: dnn_interface.h:63
ff_vf_dnn_processing
const AVFilter ff_vf_dnn_processing
Definition: vf_dnn_processing.c:450
DNN_FLOAT
@ DNN_FLOAT
Definition: dnn_interface.h:37
FF_FILTER_FORWARD_WANTED
FF_FILTER_FORWARD_WANTED(outlink, inlink)
ff_dnn_execute_model_async
DNNReturnType ff_dnn_execute_model_async(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
Definition: dnn_filter_common.c:144
internal.h
av_image_get_linesize
int av_image_get_linesize(enum AVPixelFormat pix_fmt, int width, int plane)
Compute the size of an image line with format pix_fmt and width width for the plane plane.
Definition: imgutils.c:76
activate_async
static int activate_async(AVFilterContext *filter_ctx)
Definition: vf_dnn_processing.c:349
i
int i
Definition: input.c:407
activate
static int activate(AVFilterContext *filter_ctx)
Definition: vf_dnn_processing.c:414
av_always_inline
#define av_always_inline
Definition: attributes.h:49
DnnProcessingContext::sws_uv_height
int sws_uv_height
Definition: vf_dnn_processing.c:42
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:60
DnnProcessingContext::dnnctx
DnnContext dnnctx
Definition: vf_dnn_processing.c:40
AVFilter
Filter definition.
Definition: avfilter.h:145
activate_sync
static int activate_sync(AVFilterContext *filter_ctx)
Definition: vf_dnn_processing.c:278
ret
ret
Definition: filter_design.txt:187
AV_PIX_FMT_NV12
@ AV_PIX_FMT_NV12
planar YUV 4:2:0, 12bpp, 1 plane for Y and 1 plane for the UV components, which are interleaved (firs...
Definition: pixfmt.h:89
AVFrame::height
int height
Definition: frame.h:361
sws_freeContext
void sws_freeContext(struct SwsContext *swsContext)
Free the swscaler context swsContext.
Definition: utils.c:2241
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:65
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:225
AV_PIX_FMT_YUV444P
@ AV_PIX_FMT_YUV444P
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
Definition: pixfmt.h:71
AVFilterContext
An instance of a filter.
Definition: avfilter.h:333
isPlanarYUV
static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt)
Definition: vf_dnn_processing.c:160
LOG_FORMAT_CHANNEL_MISMATCH
#define LOG_FORMAT_CHANNEL_MISMATCH()
Definition: vf_dnn_processing.c:82
desc
const char * desc
Definition: libsvtav1.c:79
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
AV_PIX_FMT_YUV422P
@ AV_PIX_FMT_YUV422P
planar YUV 4:2:2, 16bpp, (1 Cr & Cb sample per 2x1 Y samples)
Definition: pixfmt.h:70
DNNData::channels
int channels
Definition: dnn_interface.h:61
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: vf_dnn_processing.c:248
AVPixFmtDescriptor
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
Definition: pixdesc.h:70
ff_dnn_execute_model
DNNReturnType ff_dnn_execute_model(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
Definition: dnn_filter_common.c:132
ff_dnn_init
int ff_dnn_init(DnnContext *ctx, DNNFunctionType func_type, AVFilterContext *filter_ctx)
Definition: dnn_filter_common.c:54
DNNData::width
int width
Definition: dnn_interface.h:61
AV_PIX_FMT_YUV411P
@ AV_PIX_FMT_YUV411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples)
Definition: pixfmt.h:73
ff_dnn_flush
DNNReturnType ff_dnn_flush(DnnContext *ctx)
Definition: dnn_filter_common.c:176
imgutils.h
prepare_uv_scale
static int prepare_uv_scale(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:167
AVFrame::linesize
int linesize[AV_NUM_DATA_POINTERS]
For video, size in bytes of each picture line.
Definition: frame.h:334
AV_PIX_FMT_YUV410P
@ AV_PIX_FMT_YUV410P
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
Definition: pixfmt.h:72
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:28
FLAGS
#define FLAGS
Definition: vf_dnn_processing.c:46
ff_dnn_uninit
void ff_dnn_uninit(DnnContext *ctx)
Definition: dnn_filter_common.c:181
DAST_NOT_READY
@ DAST_NOT_READY
Definition: dnn_interface.h:48
SwsContext
Definition: swscale_internal.h:283
DNNAsyncStatusType
DNNAsyncStatusType
Definition: dnn_interface.h:45
AV_OPT_TYPE_CONST
@ AV_OPT_TYPE_CONST
Definition: opt.h:234
DFT_PROCESS_FRAME
@ DFT_PROCESS_FRAME
Definition: dnn_interface.h:54
ff_dnn_get_async_result
DNNAsyncStatusType ff_dnn_get_async_result(DnnContext *ctx, AVFrame **in_frame, AVFrame **out_frame)
Definition: dnn_filter_common.c:171
DnnProcessingContext::sws_uv_scale
struct SwsContext * sws_uv_scale
Definition: vf_dnn_processing.c:41
SWS_BICUBIC
#define SWS_BICUBIC
Definition: swscale.h:60
swscale.h
dnn_processing_outputs
static const AVFilterPad dnn_processing_outputs[]
Definition: vf_dnn_processing.c:441
av_get_pix_fmt_name
const char * av_get_pix_fmt_name(enum AVPixelFormat pix_fmt)
Return the short name for a pixel format, NULL in case pix_fmt is unknown.
Definition: pixdesc.c:2461
config_output
static int config_output(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:198