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 
68 static const enum AVPixelFormat pix_fmts[] = {
75 };
76 
77 #define LOG_FORMAT_CHANNEL_MISMATCH() \
78  av_log(ctx, AV_LOG_ERROR, \
79  "the frame's format %s does not match " \
80  "the model input channel %d\n", \
81  av_get_pix_fmt_name(fmt), \
82  model_input->channels);
83 
84 static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
85 {
86  AVFilterContext *ctx = inlink->dst;
87  enum AVPixelFormat fmt = inlink->format;
88 
89  // the design is to add explicit scale filter before this filter
90  if (model_input->height != -1 && model_input->height != inlink->h) {
91  av_log(ctx, AV_LOG_ERROR, "the model requires frame height %d but got %d\n",
92  model_input->height, inlink->h);
93  return AVERROR(EIO);
94  }
95  if (model_input->width != -1 && model_input->width != inlink->w) {
96  av_log(ctx, AV_LOG_ERROR, "the model requires frame width %d but got %d\n",
97  model_input->width, inlink->w);
98  return AVERROR(EIO);
99  }
100  if (model_input->dt != DNN_FLOAT) {
101  avpriv_report_missing_feature(ctx, "data type rather than DNN_FLOAT");
102  return AVERROR(EIO);
103  }
104 
105  switch (fmt) {
106  case AV_PIX_FMT_RGB24:
107  case AV_PIX_FMT_BGR24:
108  if (model_input->channels != 3) {
110  return AVERROR(EIO);
111  }
112  return 0;
113  case AV_PIX_FMT_GRAYF32:
114  case AV_PIX_FMT_YUV420P:
115  case AV_PIX_FMT_YUV422P:
116  case AV_PIX_FMT_YUV444P:
117  case AV_PIX_FMT_YUV410P:
118  case AV_PIX_FMT_YUV411P:
119  case AV_PIX_FMT_NV12:
120  if (model_input->channels != 1) {
122  return AVERROR(EIO);
123  }
124  return 0;
125  default:
127  return AVERROR(EIO);
128  }
129 
130  return 0;
131 }
132 
134 {
138  DNNData model_input;
139  int check;
140 
141  result = ff_dnn_get_input(&ctx->dnnctx, &model_input);
142  if (result != DNN_SUCCESS) {
143  av_log(ctx, AV_LOG_ERROR, "could not get input from the model\n");
144  return AVERROR(EIO);
145  }
146 
147  check = check_modelinput_inlink(&model_input, inlink);
148  if (check != 0) {
149  return check;
150  }
151 
152  return 0;
153 }
154 
156 {
158  av_assert0(desc);
159  return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components == 3;
160 }
161 
162 static int prepare_uv_scale(AVFilterLink *outlink)
163 {
164  AVFilterContext *context = outlink->src;
166  AVFilterLink *inlink = context->inputs[0];
167  enum AVPixelFormat fmt = inlink->format;
168 
169  if (isPlanarYUV(fmt)) {
170  if (inlink->w != outlink->w || inlink->h != outlink->h) {
171  if (fmt == AV_PIX_FMT_NV12) {
172  ctx->sws_uv_scale = sws_getContext(inlink->w >> 1, inlink->h >> 1, AV_PIX_FMT_YA8,
173  outlink->w >> 1, outlink->h >> 1, AV_PIX_FMT_YA8,
175  ctx->sws_uv_height = inlink->h >> 1;
176  } else {
178  int sws_src_h = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
179  int sws_src_w = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
180  int sws_dst_h = AV_CEIL_RSHIFT(outlink->h, desc->log2_chroma_h);
181  int sws_dst_w = AV_CEIL_RSHIFT(outlink->w, desc->log2_chroma_w);
182  ctx->sws_uv_scale = sws_getContext(sws_src_w, sws_src_h, AV_PIX_FMT_GRAY8,
183  sws_dst_w, sws_dst_h, AV_PIX_FMT_GRAY8,
185  ctx->sws_uv_height = sws_src_h;
186  }
187  }
188  }
189 
190  return 0;
191 }
192 
193 static int config_output(AVFilterLink *outlink)
194 {
195  AVFilterContext *context = outlink->src;
198  AVFilterLink *inlink = context->inputs[0];
199 
200  // have a try run in case that the dnn model resize the frame
201  result = ff_dnn_get_output(&ctx->dnnctx, inlink->w, inlink->h, &outlink->w, &outlink->h);
202  if (result != DNN_SUCCESS) {
203  av_log(ctx, AV_LOG_ERROR, "could not get output from the model\n");
204  return AVERROR(EIO);
205  }
206 
207  prepare_uv_scale(outlink);
208 
209  return 0;
210 }
211 
213 {
214  const AVPixFmtDescriptor *desc;
215  int uv_height;
216 
217  if (!ctx->sws_uv_scale) {
218  av_assert0(in->height == out->height && in->width == out->width);
220  uv_height = AV_CEIL_RSHIFT(in->height, desc->log2_chroma_h);
221  for (int i = 1; i < 3; ++i) {
222  int bytewidth = av_image_get_linesize(in->format, in->width, i);
223  if (bytewidth < 0) {
224  return AVERROR(EINVAL);
225  }
226  av_image_copy_plane(out->data[i], out->linesize[i],
227  in->data[i], in->linesize[i],
228  bytewidth, uv_height);
229  }
230  } else if (in->format == AV_PIX_FMT_NV12) {
231  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
232  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
233  } else {
234  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
235  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
236  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 2), in->linesize + 2,
237  0, ctx->sws_uv_height, out->data + 2, out->linesize + 2);
238  }
239 
240  return 0;
241 }
242 
243 static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
244 {
245  DnnProcessingContext *ctx = outlink->src->priv;
246  int ret;
247  DNNAsyncStatusType async_state;
248 
249  ret = ff_dnn_flush(&ctx->dnnctx);
250  if (ret != DNN_SUCCESS) {
251  return -1;
252  }
253 
254  do {
255  AVFrame *in_frame = NULL;
256  AVFrame *out_frame = NULL;
257  async_state = ff_dnn_get_result(&ctx->dnnctx, &in_frame, &out_frame);
258  if (out_frame) {
259  if (isPlanarYUV(in_frame->format))
260  copy_uv_planes(ctx, out_frame, in_frame);
261  av_frame_free(&in_frame);
262  ret = ff_filter_frame(outlink, out_frame);
263  if (ret < 0)
264  return ret;
265  if (out_pts)
266  *out_pts = out_frame->pts + pts;
267  }
268  av_usleep(5000);
269  } while (async_state >= DAST_NOT_READY);
270 
271  return 0;
272 }
273 
275 {
276  AVFilterLink *inlink = filter_ctx->inputs[0];
277  AVFilterLink *outlink = filter_ctx->outputs[0];
279  AVFrame *in = NULL, *out = NULL;
280  int64_t pts;
281  int ret, status;
282  int got_frame = 0;
283  int async_state;
284 
286 
287  do {
288  // drain all input frames
290  if (ret < 0)
291  return ret;
292  if (ret > 0) {
293  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
294  if (!out) {
295  av_frame_free(&in);
296  return AVERROR(ENOMEM);
297  }
299  if (ff_dnn_execute_model(&ctx->dnnctx, in, out) != DNN_SUCCESS) {
300  return AVERROR(EIO);
301  }
302  }
303  } while (ret > 0);
304 
305  // drain all processed frames
306  do {
307  AVFrame *in_frame = NULL;
308  AVFrame *out_frame = NULL;
309  async_state = ff_dnn_get_result(&ctx->dnnctx, &in_frame, &out_frame);
310  if (out_frame) {
311  if (isPlanarYUV(in_frame->format))
312  copy_uv_planes(ctx, out_frame, in_frame);
313  av_frame_free(&in_frame);
314  ret = ff_filter_frame(outlink, out_frame);
315  if (ret < 0)
316  return ret;
317  got_frame = 1;
318  }
319  } while (async_state == DAST_SUCCESS);
320 
321  // if frame got, schedule to next filter
322  if (got_frame)
323  return 0;
324 
326  if (status == AVERROR_EOF) {
327  int64_t out_pts = pts;
328  ret = flush_frame(outlink, pts, &out_pts);
329  ff_outlink_set_status(outlink, status, out_pts);
330  return ret;
331  }
332  }
333 
335 
336  return 0;
337 }
338 
340 {
342 
343  sws_freeContext(context->sws_uv_scale);
344  ff_dnn_uninit(&context->dnnctx);
345 }
346 
348  {
349  .name = "default",
350  .type = AVMEDIA_TYPE_VIDEO,
351  .config_props = config_input,
352  },
353 };
354 
356  {
357  .name = "default",
358  .type = AVMEDIA_TYPE_VIDEO,
359  .config_props = config_output,
360  },
361 };
362 
364  .name = "dnn_processing",
365  .description = NULL_IF_CONFIG_SMALL("Apply DNN processing filter to the input."),
366  .priv_size = sizeof(DnnProcessingContext),
367  .init = init,
368  .uninit = uninit,
372  .priv_class = &dnn_processing_class,
373  .activate = activate,
374 };
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:98
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_dnn_processing.c:339
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:49
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:114
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:1018
av_pix_fmt_desc_get
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:2564
AVERROR_EOF
#define AVERROR_EOF
End of file.
Definition: error.h:57
FILTER_PIXFMTS_ARRAY
#define FILTER_PIXFMTS_ARRAY(array)
Definition: internal.h:171
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:112
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:247
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:109
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:1204
DNNData::height
int height
Definition: dnn_interface.h:61
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_dnn_processing.c:133
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:153
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
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:1417
AVFilterContext::priv
void * priv
private data for use by the filter
Definition: avfilter.h:401
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:653
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:50
flush_frame
static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
Definition: vf_dnn_processing.c:243
avassert.h
AV_LOG_ERROR
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:180
av_cold
#define av_cold
Definition: attributes.h:90
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:212
AV_CEIL_RSHIFT
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:51
pix_fmts
static enum AVPixelFormat pix_fmts[]
Definition: vf_dnn_processing.c:68
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
filters.h
ff_dnn_get_result
DNNAsyncStatusType ff_dnn_get_result(DnnContext *ctx, AVFrame **in_frame, AVFrame **out_frame)
Definition: dnn_filter_common.c:147
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:423
check_modelinput_inlink
static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
Definition: vf_dnn_processing.c:84
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:191
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:66
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:537
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:347
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:1371
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:136
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:2007
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:363
DNN_FLOAT
@ DNN_FLOAT
Definition: dnn_interface.h:37
FF_FILTER_FORWARD_WANTED
FF_FILTER_FORWARD_WANTED(outlink, inlink)
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
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:271
activate
static int activate(AVFilterContext *filter_ctx)
Definition: vf_dnn_processing.c:274
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:56
DnnProcessingContext::dnnctx
DnnContext dnnctx
Definition: vf_dnn_processing.c:40
AVFilter
Filter definition.
Definition: avfilter.h:149
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:2338
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:65
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:224
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:386
isPlanarYUV
static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt)
Definition: vf_dnn_processing.c:155
LOG_FORMAT_CHANNEL_MISMATCH
#define LOG_FORMAT_CHANNEL_MISMATCH()
Definition: vf_dnn_processing.c:77
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
AVPixFmtDescriptor
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
Definition: pixdesc.h:69
ff_dnn_execute_model
DNNReturnType ff_dnn_execute_model(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
Definition: dnn_filter_common.c:120
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:192
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:152
imgutils.h
prepare_uv_scale
static int prepare_uv_scale(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:162
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:157
DAST_NOT_READY
@ DAST_NOT_READY
Definition: dnn_interface.h:48
SwsContext
Definition: swscale_internal.h:300
DNNAsyncStatusType
DNNAsyncStatusType
Definition: dnn_interface.h:45
AV_OPT_TYPE_CONST
@ AV_OPT_TYPE_CONST
Definition: opt.h:233
DFT_PROCESS_FRAME
@ DFT_PROCESS_FRAME
Definition: dnn_interface.h:54
DnnProcessingContext::sws_uv_scale
struct SwsContext * sws_uv_scale
Definition: vf_dnn_processing.c:41
SWS_BICUBIC
#define SWS_BICUBIC
Definition: swscale.h:61
swscale.h
dnn_processing_outputs
static const AVFilterPad dnn_processing_outputs[]
Definition: vf_dnn_processing.c:355
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:2484
config_output
static int config_output(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:193