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 {
64  DnnProcessingContext *ctx = context->priv;
65  return ff_dnn_init(&ctx->dnnctx, DFT_PROCESS_FRAME, context);
66 }
67 
69 {
70  static const enum AVPixelFormat pix_fmts[] = {
77  };
78  AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
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 {
140  AVFilterContext *context = inlink->dst;
141  DnnProcessingContext *ctx = context->priv;
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 {
162  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
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;
170  DnnProcessingContext *ctx = context->priv;
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;
201  DnnProcessingContext *ctx = context->priv;
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);
224  desc = av_pix_fmt_desc_get(in->format);
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  av_image_copy_plane(out->data[i], out->linesize[i],
229  in->data[i], in->linesize[i],
230  bytewidth, uv_height);
231  }
232  } else if (in->format == AV_PIX_FMT_NV12) {
233  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
234  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
235  } else {
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  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 2), in->linesize + 2,
239  0, ctx->sws_uv_height, out->data + 2, out->linesize + 2);
240  }
241 
242  return 0;
243 }
244 
246 {
247  AVFilterContext *context = inlink->dst;
248  AVFilterLink *outlink = context->outputs[0];
249  DnnProcessingContext *ctx = context->priv;
250  DNNReturnType dnn_result;
251  AVFrame *out;
252 
253  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
254  if (!out) {
255  av_frame_free(&in);
256  return AVERROR(ENOMEM);
257  }
258  av_frame_copy_props(out, in);
259 
260  dnn_result = ff_dnn_execute_model(&ctx->dnnctx, in, out);
261  if (dnn_result != DNN_SUCCESS){
262  av_log(ctx, AV_LOG_ERROR, "failed to execute model\n");
263  av_frame_free(&in);
264  av_frame_free(&out);
265  return AVERROR(EIO);
266  }
267 
268  if (isPlanarYUV(in->format))
269  copy_uv_planes(ctx, out, in);
270 
271  av_frame_free(&in);
272  return ff_filter_frame(outlink, out);
273 }
274 
276 {
277  AVFilterLink *inlink = filter_ctx->inputs[0];
278  AVFilterLink *outlink = filter_ctx->outputs[0];
279  AVFrame *in = NULL;
280  int64_t pts;
281  int ret, status;
282  int got_frame = 0;
283 
284  FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
285 
286  do {
287  // drain all input frames
288  ret = ff_inlink_consume_frame(inlink, &in);
289  if (ret < 0)
290  return ret;
291  if (ret > 0) {
292  ret = filter_frame(inlink, in);
293  if (ret < 0)
294  return ret;
295  got_frame = 1;
296  }
297  } while (ret > 0);
298 
299  // if frame got, schedule to next filter
300  if (got_frame)
301  return 0;
302 
303  if (ff_inlink_acknowledge_status(inlink, &status, &pts)) {
304  if (status == AVERROR_EOF) {
305  ff_outlink_set_status(outlink, status, pts);
306  return ret;
307  }
308  }
309 
310  FF_FILTER_FORWARD_WANTED(outlink, inlink);
311 
312  return FFERROR_NOT_READY;
313 }
314 
315 static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
316 {
317  DnnProcessingContext *ctx = outlink->src->priv;
318  int ret;
319  DNNAsyncStatusType async_state;
320 
321  ret = ff_dnn_flush(&ctx->dnnctx);
322  if (ret != DNN_SUCCESS) {
323  return -1;
324  }
325 
326  do {
327  AVFrame *in_frame = NULL;
328  AVFrame *out_frame = NULL;
329  async_state = ff_dnn_get_async_result(&ctx->dnnctx, &in_frame, &out_frame);
330  if (out_frame) {
331  if (isPlanarYUV(in_frame->format))
332  copy_uv_planes(ctx, out_frame, in_frame);
333  av_frame_free(&in_frame);
334  ret = ff_filter_frame(outlink, out_frame);
335  if (ret < 0)
336  return ret;
337  if (out_pts)
338  *out_pts = out_frame->pts + pts;
339  }
340  av_usleep(5000);
341  } while (async_state >= DAST_NOT_READY);
342 
343  return 0;
344 }
345 
347 {
348  AVFilterLink *inlink = filter_ctx->inputs[0];
349  AVFilterLink *outlink = filter_ctx->outputs[0];
350  DnnProcessingContext *ctx = filter_ctx->priv;
351  AVFrame *in = NULL, *out = NULL;
352  int64_t pts;
353  int ret, status;
354  int got_frame = 0;
355  int async_state;
356 
357  FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
358 
359  do {
360  // drain all input frames
361  ret = ff_inlink_consume_frame(inlink, &in);
362  if (ret < 0)
363  return ret;
364  if (ret > 0) {
365  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
366  if (!out) {
367  av_frame_free(&in);
368  return AVERROR(ENOMEM);
369  }
370  av_frame_copy_props(out, in);
371  if (ff_dnn_execute_model_async(&ctx->dnnctx, in, out) != DNN_SUCCESS) {
372  return AVERROR(EIO);
373  }
374  }
375  } while (ret > 0);
376 
377  // drain all processed frames
378  do {
379  AVFrame *in_frame = NULL;
380  AVFrame *out_frame = NULL;
381  async_state = ff_dnn_get_async_result(&ctx->dnnctx, &in_frame, &out_frame);
382  if (out_frame) {
383  if (isPlanarYUV(in_frame->format))
384  copy_uv_planes(ctx, out_frame, in_frame);
385  av_frame_free(&in_frame);
386  ret = ff_filter_frame(outlink, out_frame);
387  if (ret < 0)
388  return ret;
389  got_frame = 1;
390  }
391  } while (async_state == DAST_SUCCESS);
392 
393  // if frame got, schedule to next filter
394  if (got_frame)
395  return 0;
396 
397  if (ff_inlink_acknowledge_status(inlink, &status, &pts)) {
398  if (status == AVERROR_EOF) {
399  int64_t out_pts = pts;
400  ret = flush_frame(outlink, pts, &out_pts);
401  ff_outlink_set_status(outlink, status, out_pts);
402  return ret;
403  }
404  }
405 
406  FF_FILTER_FORWARD_WANTED(outlink, inlink);
407 
408  return 0;
409 }
410 
412 {
413  DnnProcessingContext *ctx = filter_ctx->priv;
414 
415  if (ctx->dnnctx.async)
416  return activate_async(filter_ctx);
417  else
418  return activate_sync(filter_ctx);
419 }
420 
422 {
424 
425  sws_freeContext(context->sws_uv_scale);
426  ff_dnn_uninit(&context->dnnctx);
427 }
428 
430  {
431  .name = "default",
432  .type = AVMEDIA_TYPE_VIDEO,
433  .config_props = config_input,
434  },
435  { NULL }
436 };
437 
439  {
440  .name = "default",
441  .type = AVMEDIA_TYPE_VIDEO,
442  .config_props = config_output,
443  },
444  { NULL }
445 };
446 
448  .name = "dnn_processing",
449  .description = NULL_IF_CONFIG_SMALL("Apply DNN processing filter to the input."),
450  .priv_size = sizeof(DnnProcessingContext),
451  .init = init,
452  .uninit = uninit,
454  .inputs = dnn_processing_inputs,
455  .outputs = dnn_processing_outputs,
456  .priv_class = &dnn_processing_class,
457  .activate = activate,
458 };
DNNReturnType ff_dnn_get_output(DnnContext *ctx, int input_width, int input_height, int *output_width, int *output_height)
int ff_inlink_consume_frame(AVFilterLink *link, AVFrame **rframe)
Take a frame from the link&#39;s FIFO and update the link&#39;s stats.
Definition: avfilter.c:1492
#define NULL
Definition: coverity.c:32
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
static enum AVPixelFormat pix_fmt
Buffered I/O operations.
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:2573
This structure describes decoded (raw) audio or video data.
Definition: frame.h:318
#define SWS_BICUBIC
Definition: swscale.h:60
AVOption.
Definition: opt.h:248
static int prepare_uv_scale(AVFilterLink *outlink)
static const AVOption dnn_processing_options[]
8 bits gray, 8 bits alpha
Definition: pixfmt.h:143
const char * desc
Definition: libsvtav1.c:79
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
Definition: pixfmt.h:71
misc image utilities
packed RGB 8:8:8, 24bpp, RGBRGB...
Definition: pixfmt.h:68
struct SwsContext * sws_uv_scale
#define OFFSET(x)
int channels
Definition: dnn_interface.h:60
DNNReturnType ff_dnn_get_input(DnnContext *ctx, DNNData *input)
static const AVFilterPad dnn_processing_outputs[]
static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
int av_usleep(unsigned usec)
Sleep for a period of time.
Definition: time.c:84
return FFERROR_NOT_READY
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
static av_cold void uninit(AVFilterContext *ctx)
uint8_t log2_chroma_w
Amount to shift the luma width right to find the chroma width.
Definition: pixdesc.h:92
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
AVFilterFormats * ff_make_format_list(const int *fmts)
Create a list of supported formats.
Definition: formats.c:287
const char * name
Pad name.
Definition: internal.h:60
AVFilterLink ** inputs
array of pointers to input links
Definition: avfilter.h:349
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
AVFilter ff_vf_dnn_processing
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1094
uint8_t
#define av_cold
Definition: attributes.h:88
AVOptions.
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:411
static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
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:1917
#define AVERROR_EOF
End of file.
Definition: error.h:55
static int config_input(AVFilterLink *inlink)
external API header
#define av_log(a,...)
#define FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink)
Forward the status on an output link to an input link.
Definition: filters.h:199
A filter pad used for either input or output.
Definition: internal.h:54
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:1447
int width
Definition: frame.h:376
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:194
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:588
uint8_t log2_chroma_h
Amount to shift the luma height right to find the chroma height.
Definition: pixdesc.h:101
int height
Definition: dnn_interface.h:60
static av_cold int init(AVFilterContext *context)
#define AV_PIX_FMT_FLAG_RGB
The pixel format contains RGB-like data (as opposed to YUV/grayscale).
Definition: pixdesc.h:148
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:204
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:117
void * priv
private data for use by the filter
Definition: avfilter.h:356
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
simple assert() macros that are a bit more flexible than ISO C assert().
static FilteringContext * filter_ctx
Definition: transcoding.c:48
planar YUV 4:2:2, 16bpp, (1 Cr & Cb sample per 2x1 Y samples)
Definition: pixfmt.h:70
uint64_t flags
Combination of AV_PIX_FMT_FLAG_...
Definition: pixdesc.h:106
uint8_t nb_components
The number of components each pixel has, (1-4)
Definition: pixdesc.h:83
DNNAsyncStatusType
Definition: dnn_interface.h:44
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
static int copy_uv_planes(DnnProcessingContext *ctx, AVFrame *out, const AVFrame *in)
AVFormatContext * ctx
Definition: movenc.c:48
void sws_freeContext(struct SwsContext *swsContext)
Free the swscaler context swsContext.
Definition: utils.c:2337
packed RGB 8:8:8, 24bpp, BGRBGR...
Definition: pixfmt.h:69
static const AVFilterPad dnn_processing_inputs[]
AVFILTER_DEFINE_CLASS(dnn_processing)
#define DNN_COMMON_OPTIONS
DNNReturnType
Definition: dnn_interface.h:33
static const AVFilterPad outputs[]
Definition: af_acontrast.c:203
int format
format of the frame, -1 if unknown or unset Values correspond to enum AVPixelFormat for video frames...
Definition: frame.h:391
static int activate(AVFilterContext *filter_ctx)
#define AV_PIX_FMT_GRAYF32
Definition: pixfmt.h:431
int linesize[AV_NUM_DATA_POINTERS]
For video, size in bytes of each picture line.
Definition: frame.h:349
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
Definition: pixdesc.h:81
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&#39;t need to export the SwsContext.
Definition: swscale.c:745
common functions for the dnn based filters
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(const int16_t *) pi >> 8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(const int16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(const int16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(const int32_t *) pi >> 24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(const int32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(const int32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(const float *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(const float *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(const float *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(const double *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(const double *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(const double *) pi *(1U<< 31))))#define SET_CONV_FUNC_GROUP(ofmt, ifmt) static void set_generic_function(AudioConvert *ac){}void ff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, int channels, int sample_rate, int apply_map){AudioConvert *ac;int in_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) return NULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt) > 2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);return NULL;}return ac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}else if(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;else ac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);return ac;}int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){int use_generic=1;int len=in->nb_samples;int p;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%d samples - audio_convert: %s to %s (dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));return ff_convert_dither(ac-> in
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
Definition: pixfmt.h:72
Describe the class of an AVClass context structure.
Definition: log.h:67
Filter definition.
Definition: avfilter.h:145
static int query_formats(AVFilterContext *context)
static int activate_async(AVFilterContext *filter_ctx)
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
static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt)
static int activate_sync(AVFilterContext *filter_ctx)
const char * name
Filter name.
Definition: avfilter.h:149
AVFilterLink ** outputs
array of pointers to output links
Definition: avfilter.h:353
static enum AVPixelFormat pix_fmts[]
Definition: libkvazaar.c:303
void avpriv_report_missing_feature(void *avc, const char *msg,...) av_printf_format(2
Log a generic warning message about a missing feature.
static int64_t pts
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:332
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
DNNReturnType ff_dnn_execute_model_async(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
Definition: pixfmt.h:66
Y , 8bpp.
Definition: pixfmt.h:74
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
DNNReturnType ff_dnn_flush(DnnContext *ctx)
void ff_dnn_uninit(DnnContext *ctx)
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples)
Definition: pixfmt.h:73
FF_FILTER_FORWARD_WANTED(outlink, inlink)
DNNReturnType ff_dnn_execute_model(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
DNNAsyncStatusType ff_dnn_get_async_result(DnnContext *ctx, AVFrame **in_frame, AVFrame **out_frame)
#define LOG_FORMAT_CHANNEL_MISMATCH()
A list of supported formats for one end of a filter link.
Definition: formats.h:65
An instance of a filter.
Definition: avfilter.h:341
and forward the result(frame or status change) to the corresponding input.If nothing is possible
int height
Definition: frame.h:376
FILE * out
Definition: movenc.c:54
#define av_always_inline
Definition: attributes.h:45
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:373
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:2489
internal API functions
static int config_output(AVFilterLink *outlink)
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
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 local context
DNNDataType dt
Definition: dnn_interface.h:62
AVPixelFormat
Pixel format.
Definition: pixfmt.h:64
int av_frame_copy_props(AVFrame *dst, const AVFrame *src)
Copy only "metadata" fields from src to dst.
Definition: frame.c:659
int i
Definition: input.c:407
int ff_dnn_init(DnnContext *ctx, DNNFunctionType func_type, AVFilterContext *filter_ctx)
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:58
#define FLAGS
#define check(x, y, S, v)