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 "libavutil/opt.h"
27 #include "libavutil/pixdesc.h"
28 #include "libavutil/avassert.h"
29 #include "libavutil/imgutils.h"
30 #include "filters.h"
31 #include "dnn_filter_common.h"
32 #include "internal.h"
33 #include "video.h"
34 #include "libswscale/swscale.h"
35 #include "libavutil/time.h"
36 
37 typedef struct DnnProcessingContext {
38  const AVClass *class;
43 
44 #define OFFSET(x) offsetof(DnnProcessingContext, dnnctx.x)
45 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM
46 static const AVOption dnn_processing_options[] = {
47  { "dnn_backend", "DNN backend", OFFSET(backend_type), AV_OPT_TYPE_INT, { .i64 = DNN_TF }, INT_MIN, INT_MAX, FLAGS, .unit = "backend" },
48 #if (CONFIG_LIBTENSORFLOW == 1)
49  { "tensorflow", "tensorflow backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = DNN_TF }, 0, 0, FLAGS, .unit = "backend" },
50 #endif
51 #if (CONFIG_LIBOPENVINO == 1)
52  { "openvino", "openvino backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = DNN_OV }, 0, 0, FLAGS, .unit = "backend" },
53 #endif
54 #if (CONFIG_LIBTORCH == 1)
55  { "torch", "torch backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = DNN_TH }, 0, 0, FLAGS, "backend" },
56 #endif
58  { NULL }
59 };
60 
61 AVFILTER_DEFINE_CLASS(dnn_processing);
62 
64 {
66  return ff_dnn_init(&ctx->dnnctx, DFT_PROCESS_FRAME, context);
67 }
68 
69 static const enum AVPixelFormat pix_fmts[] = {
76 };
77 
78 #define LOG_FORMAT_CHANNEL_MISMATCH() \
79  av_log(ctx, AV_LOG_ERROR, \
80  "the frame's format %s does not match " \
81  "the model input channel %d\n", \
82  av_get_pix_fmt_name(fmt), \
83  model_input->dims[dnn_get_channel_idx_by_layout(model_input->layout)]);
84 
85 static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
86 {
87  AVFilterContext *ctx = inlink->dst;
88  enum AVPixelFormat fmt = inlink->format;
89  int width_idx, height_idx;
90 
91  width_idx = dnn_get_width_idx_by_layout(model_input->layout);
92  height_idx = dnn_get_height_idx_by_layout(model_input->layout);
93  // the design is to add explicit scale filter before this filter
94  if (model_input->dims[height_idx] != -1 &&
95  model_input->dims[height_idx] != inlink->h) {
96  av_log(ctx, AV_LOG_ERROR, "the model requires frame height %d but got %d\n",
97  model_input->dims[height_idx],
98  inlink->h);
99  return AVERROR(EIO);
100  }
101  if (model_input->dims[width_idx] != -1 &&
102  model_input->dims[width_idx] != inlink->w) {
103  av_log(ctx, AV_LOG_ERROR, "the model requires frame width %d but got %d\n",
104  model_input->dims[width_idx],
105  inlink->w);
106  return AVERROR(EIO);
107  }
108  if (model_input->dt != DNN_FLOAT) {
109  avpriv_report_missing_feature(ctx, "data type rather than DNN_FLOAT");
110  return AVERROR(EIO);
111  }
112 
113  switch (fmt) {
114  case AV_PIX_FMT_RGB24:
115  case AV_PIX_FMT_BGR24:
116  if (model_input->dims[dnn_get_channel_idx_by_layout(model_input->layout)] != 3) {
118  return AVERROR(EIO);
119  }
120  return 0;
121  case AV_PIX_FMT_GRAY8:
122  case AV_PIX_FMT_GRAYF32:
123  case AV_PIX_FMT_YUV420P:
124  case AV_PIX_FMT_YUV422P:
125  case AV_PIX_FMT_YUV444P:
126  case AV_PIX_FMT_YUV410P:
127  case AV_PIX_FMT_YUV411P:
128  case AV_PIX_FMT_NV12:
129  if (model_input->dims[dnn_get_channel_idx_by_layout(model_input->layout)] != 1) {
131  return AVERROR(EIO);
132  }
133  return 0;
134  default:
136  return AVERROR(EIO);
137  }
138 
139  return 0;
140 }
141 
143 {
146  int result;
147  DNNData model_input;
148  int check;
149 
150  result = ff_dnn_get_input(&ctx->dnnctx, &model_input);
151  if (result != 0) {
152  av_log(ctx, AV_LOG_ERROR, "could not get input from the model\n");
153  return result;
154  }
155 
156  check = check_modelinput_inlink(&model_input, inlink);
157  if (check != 0) {
158  return check;
159  }
160 
161  return 0;
162 }
163 
165 {
167  av_assert0(desc);
168  return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components == 3;
169 }
170 
171 static int prepare_uv_scale(AVFilterLink *outlink)
172 {
173  AVFilterContext *context = outlink->src;
175  AVFilterLink *inlink = context->inputs[0];
176  enum AVPixelFormat fmt = inlink->format;
177 
178  if (isPlanarYUV(fmt)) {
179  if (inlink->w != outlink->w || inlink->h != outlink->h) {
180  if (fmt == AV_PIX_FMT_NV12) {
181  ctx->sws_uv_scale = sws_getContext(inlink->w >> 1, inlink->h >> 1, AV_PIX_FMT_YA8,
182  outlink->w >> 1, outlink->h >> 1, AV_PIX_FMT_YA8,
184  ctx->sws_uv_height = inlink->h >> 1;
185  } else {
187  int sws_src_h = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
188  int sws_src_w = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
189  int sws_dst_h = AV_CEIL_RSHIFT(outlink->h, desc->log2_chroma_h);
190  int sws_dst_w = AV_CEIL_RSHIFT(outlink->w, desc->log2_chroma_w);
191  ctx->sws_uv_scale = sws_getContext(sws_src_w, sws_src_h, AV_PIX_FMT_GRAY8,
192  sws_dst_w, sws_dst_h, AV_PIX_FMT_GRAY8,
194  ctx->sws_uv_height = sws_src_h;
195  }
196  }
197  }
198 
199  return 0;
200 }
201 
202 static int config_output(AVFilterLink *outlink)
203 {
204  AVFilterContext *context = outlink->src;
206  int result;
207  AVFilterLink *inlink = context->inputs[0];
208 
209  // have a try run in case that the dnn model resize the frame
210  result = ff_dnn_get_output(&ctx->dnnctx, inlink->w, inlink->h, &outlink->w, &outlink->h);
211  if (result != 0) {
212  av_log(ctx, AV_LOG_ERROR, "could not get output from the model\n");
213  return result;
214  }
215 
216  prepare_uv_scale(outlink);
217 
218  return 0;
219 }
220 
222 {
223  const AVPixFmtDescriptor *desc;
224  int uv_height;
225 
226  if (!ctx->sws_uv_scale) {
227  av_assert0(in->height == out->height && in->width == out->width);
229  uv_height = AV_CEIL_RSHIFT(in->height, desc->log2_chroma_h);
230  for (int i = 1; i < 3; ++i) {
231  int bytewidth = av_image_get_linesize(in->format, in->width, i);
232  if (bytewidth < 0) {
233  return AVERROR(EINVAL);
234  }
235  av_image_copy_plane(out->data[i], out->linesize[i],
236  in->data[i], in->linesize[i],
237  bytewidth, uv_height);
238  }
239  } else if (in->format == AV_PIX_FMT_NV12) {
240  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
241  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
242  } else {
243  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
244  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
245  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 2), in->linesize + 2,
246  0, ctx->sws_uv_height, out->data + 2, out->linesize + 2);
247  }
248 
249  return 0;
250 }
251 
252 static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
253 {
254  DnnProcessingContext *ctx = outlink->src->priv;
255  int ret;
256  DNNAsyncStatusType async_state;
257 
258  ret = ff_dnn_flush(&ctx->dnnctx);
259  if (ret != 0) {
260  return -1;
261  }
262 
263  do {
264  AVFrame *in_frame = NULL;
265  AVFrame *out_frame = NULL;
266  async_state = ff_dnn_get_result(&ctx->dnnctx, &in_frame, &out_frame);
267  if (out_frame) {
268  if (isPlanarYUV(in_frame->format))
269  copy_uv_planes(ctx, out_frame, in_frame);
270  av_frame_free(&in_frame);
271  ret = ff_filter_frame(outlink, out_frame);
272  if (ret < 0)
273  return ret;
274  if (out_pts)
275  *out_pts = out_frame->pts + pts;
276  }
277  av_usleep(5000);
278  } while (async_state >= DAST_NOT_READY);
279 
280  return 0;
281 }
282 
284 {
285  AVFilterLink *inlink = filter_ctx->inputs[0];
286  AVFilterLink *outlink = filter_ctx->outputs[0];
288  AVFrame *in = NULL, *out = NULL;
289  int64_t pts;
290  int ret, status;
291  int got_frame = 0;
292  int async_state;
293 
295 
296  do {
297  // drain all input frames
299  if (ret < 0)
300  return ret;
301  if (ret > 0) {
302  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
303  if (!out) {
304  av_frame_free(&in);
305  return AVERROR(ENOMEM);
306  }
308  if (ff_dnn_execute_model(&ctx->dnnctx, in, out) != 0) {
309  return AVERROR(EIO);
310  }
311  }
312  } while (ret > 0);
313 
314  // drain all processed frames
315  do {
316  AVFrame *in_frame = NULL;
317  AVFrame *out_frame = NULL;
318  async_state = ff_dnn_get_result(&ctx->dnnctx, &in_frame, &out_frame);
319  if (out_frame) {
320  if (isPlanarYUV(in_frame->format))
321  copy_uv_planes(ctx, out_frame, in_frame);
322  av_frame_free(&in_frame);
323  ret = ff_filter_frame(outlink, out_frame);
324  if (ret < 0)
325  return ret;
326  got_frame = 1;
327  }
328  } while (async_state == DAST_SUCCESS);
329 
330  // if frame got, schedule to next filter
331  if (got_frame)
332  return 0;
333 
335  if (status == AVERROR_EOF) {
336  int64_t out_pts = pts;
337  ret = flush_frame(outlink, pts, &out_pts);
338  ff_outlink_set_status(outlink, status, out_pts);
339  return ret;
340  }
341  }
342 
344 
345  return 0;
346 }
347 
349 {
351 
352  sws_freeContext(context->sws_uv_scale);
353  ff_dnn_uninit(&context->dnnctx);
354 }
355 
357  {
358  .name = "default",
359  .type = AVMEDIA_TYPE_VIDEO,
360  .config_props = config_input,
361  },
362 };
363 
365  {
366  .name = "default",
367  .type = AVMEDIA_TYPE_VIDEO,
368  .config_props = config_output,
369  },
370 };
371 
373  .name = "dnn_processing",
374  .description = NULL_IF_CONFIG_SMALL("Apply DNN processing filter to the input."),
375  .priv_size = sizeof(DnnProcessingContext),
376  .init = init,
377  .uninit = uninit,
381  .priv_class = &dnn_processing_class,
382  .activate = activate,
383 };
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:112
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_dnn_processing.c:348
dnn_processing_options
static const AVOption dnn_processing_options[]
Definition: vf_dnn_processing.c:46
AVPixelFormat
AVPixelFormat
Pixel format.
Definition: pixfmt.h:71
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:140
out
FILE * out
Definition: movenc.c:54
OFFSET
#define OFFSET(x)
Definition: vf_dnn_processing.c:44
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:2962
AVERROR_EOF
#define AVERROR_EOF
End of file.
Definition: error.h:57
FILTER_PIXFMTS_ARRAY
#define FILTER_PIXFMTS_ARRAY(array)
Definition: internal.h:162
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:160
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:375
pixdesc.h
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:487
DnnProcessingContext
Definition: vf_dnn_processing.c:37
AVFrame::width
int width
Definition: frame.h:447
AVOption
AVOption.
Definition: opt.h:346
AV_PIX_FMT_BGR24
@ AV_PIX_FMT_BGR24
packed RGB 8:8:8, 24bpp, BGRBGR...
Definition: pixfmt.h:76
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:1205
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_dnn_processing.c:142
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:170
dnn_filter_common.h
video.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:396
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
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:1445
dnn_get_width_idx_by_layout
static int dnn_get_width_idx_by_layout(DNNLayout layout)
Definition: dnn_interface.h:137
AVFilterContext::priv
void * priv
private data for use by the filter
Definition: avfilter.h:422
DnnContext
Definition: dnn_filter_common.h:29
filter_ctx
static FilteringContext * filter_ctx
Definition: transcode.c:51
pts
static int64_t pts
Definition: transcode_aac.c:643
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:33
flush_frame
static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
Definition: vf_dnn_processing.c:252
avassert.h
DNN_TF
@ DNN_TF
Definition: dnn_interface.h:35
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:221
AV_CEIL_RSHIFT
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:58
pix_fmt
static enum AVPixelFormat pix_fmt
Definition: demux_decode.c:41
pix_fmts
static enum AVPixelFormat pix_fmts[]
Definition: vf_dnn_processing.c:69
av_assert0
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:40
DNNData
Definition: dnn_interface.h:65
filters.h
ff_dnn_get_result
DNNAsyncStatusType ff_dnn_get_result(DnnContext *ctx, AVFrame **in_frame, AVFrame **out_frame)
Definition: dnn_filter_common.c:161
ctx
AVFormatContext * ctx
Definition: movenc.c:48
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:73
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:511
check_modelinput_inlink
static int check_modelinput_inlink(const DNNData *model_input, const AVFilterLink *inlink)
Definition: vf_dnn_processing.c:85
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:182
ff_dnn_get_input
int ff_dnn_get_input(DnnContext *ctx, DNNData *input)
Definition: dnn_filter_common.c:121
DNN_OV
@ DNN_OV
Definition: dnn_interface.h:35
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:709
init
static av_cold int init(AVFilterContext *context)
Definition: vf_dnn_processing.c:63
time.h
dnn_processing_inputs
static const AVFilterPad dnn_processing_inputs[]
Definition: vf_dnn_processing.c:356
AV_PIX_FMT_GRAY8
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Definition: pixfmt.h:81
ff_dnn_flush
int ff_dnn_flush(DnnContext *ctx)
Definition: dnn_filter_common.c:166
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:1392
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:75
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:106
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:2094
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:462
DNNData::dt
DNNDataType dt
Definition: dnn_interface.h:69
DNNData::layout
DNNLayout layout
Definition: dnn_interface.h:71
ff_vf_dnn_processing
const AVFilter ff_vf_dnn_processing
Definition: vf_dnn_processing.c:372
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:255
activate
static int activate(AVFilterContext *filter_ctx)
Definition: vf_dnn_processing.c:283
av_always_inline
#define av_always_inline
Definition: attributes.h:49
DnnProcessingContext::sws_uv_height
int sws_uv_height
Definition: vf_dnn_processing.c:41
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:39
DnnProcessingContext::dnnctx
DnnContext dnnctx
Definition: vf_dnn_processing.c:39
AVFilter
Filter definition.
Definition: avfilter.h:166
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:96
AVFrame::height
int height
Definition: frame.h:447
status
ov_status_e status
Definition: dnn_backend_openvino.c:120
sws_freeContext
void sws_freeContext(struct SwsContext *swsContext)
Free the swscaler context swsContext.
Definition: utils.c:2425
ff_dnn_get_output
int ff_dnn_get_output(DnnContext *ctx, int input_width, int input_height, int *output_width, int *output_height)
Definition: dnn_filter_common.c:126
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:72
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:235
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:78
AVFilterContext
An instance of a filter.
Definition: avfilter.h:407
isPlanarYUV
static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt)
Definition: vf_dnn_processing.c:164
LOG_FORMAT_CHANNEL_MISMATCH
#define LOG_FORMAT_CHANNEL_MISMATCH()
Definition: vf_dnn_processing.c:78
DNNData::dims
int dims[4]
Definition: dnn_interface.h:67
desc
const char * desc
Definition: libsvtav1.c:75
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
DNN_TH
@ DNN_TH
Definition: dnn_interface.h:35
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:77
dnn_get_height_idx_by_layout
static int dnn_get_height_idx_by_layout(DNNLayout layout)
Definition: dnn_interface.h:142
AVPixFmtDescriptor
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
Definition: pixdesc.h:69
dnn_get_channel_idx_by_layout
static int dnn_get_channel_idx_by_layout(DNNLayout layout)
Definition: dnn_interface.h:147
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:183
ff_dnn_init
int ff_dnn_init(DnnContext *ctx, DNNFunctionType func_type, AVFilterContext *filter_ctx)
Definition: dnn_filter_common.c:54
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:80
imgutils.h
prepare_uv_scale
static int prepare_uv_scale(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:171
AVFrame::linesize
int linesize[AV_NUM_DATA_POINTERS]
For video, a positive or negative value, which is typically indicating the size in bytes of each pict...
Definition: frame.h:420
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:79
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
FLAGS
#define FLAGS
Definition: vf_dnn_processing.c:45
ff_dnn_uninit
void ff_dnn_uninit(DnnContext *ctx)
Definition: dnn_filter_common.c:171
ff_dnn_execute_model
int ff_dnn_execute_model(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
Definition: dnn_filter_common.c:134
DAST_NOT_READY
@ DAST_NOT_READY
Definition: dnn_interface.h:48
SwsContext
Definition: swscale_internal.h:299
DNNAsyncStatusType
DNNAsyncStatusType
Definition: dnn_interface.h:45
AV_OPT_TYPE_CONST
@ AV_OPT_TYPE_CONST
Definition: opt.h:244
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:40
SWS_BICUBIC
#define SWS_BICUBIC
Definition: swscale.h:67
swscale.h
dnn_processing_outputs
static const AVFilterPad dnn_processing_outputs[]
Definition: vf_dnn_processing.c:364
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:2882
config_output
static int config_output(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:202