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
57  { NULL }
58 };
59 
60 AVFILTER_DNN_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->dims[dnn_get_channel_idx_by_layout(model_input->layout)]);
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  int width_idx, height_idx;
89 
90  width_idx = dnn_get_width_idx_by_layout(model_input->layout);
91  height_idx = dnn_get_height_idx_by_layout(model_input->layout);
92  // the design is to add explicit scale filter before this filter
93  if (model_input->dims[height_idx] != -1 &&
94  model_input->dims[height_idx] != inlink->h) {
95  av_log(ctx, AV_LOG_ERROR, "the model requires frame height %d but got %d\n",
96  model_input->dims[height_idx],
97  inlink->h);
98  return AVERROR(EIO);
99  }
100  if (model_input->dims[width_idx] != -1 &&
101  model_input->dims[width_idx] != inlink->w) {
102  av_log(ctx, AV_LOG_ERROR, "the model requires frame width %d but got %d\n",
103  model_input->dims[width_idx],
104  inlink->w);
105  return AVERROR(EIO);
106  }
107  if (model_input->dt != DNN_FLOAT) {
108  avpriv_report_missing_feature(ctx, "data type rather than DNN_FLOAT");
109  return AVERROR(EIO);
110  }
111 
112  switch (fmt) {
113  case AV_PIX_FMT_RGB24:
114  case AV_PIX_FMT_BGR24:
115  if (model_input->dims[dnn_get_channel_idx_by_layout(model_input->layout)] != 3) {
117  return AVERROR(EIO);
118  }
119  return 0;
120  case AV_PIX_FMT_GRAY8:
121  case AV_PIX_FMT_GRAYF32:
122  case AV_PIX_FMT_YUV420P:
123  case AV_PIX_FMT_YUV422P:
124  case AV_PIX_FMT_YUV444P:
125  case AV_PIX_FMT_YUV410P:
126  case AV_PIX_FMT_YUV411P:
127  case AV_PIX_FMT_NV12:
128  if (model_input->dims[dnn_get_channel_idx_by_layout(model_input->layout)] != 1) {
130  return AVERROR(EIO);
131  }
132  return 0;
133  default:
135  return AVERROR(EIO);
136  }
137 
138  return 0;
139 }
140 
142 {
145  int result;
146  DNNData model_input;
147  int check;
148 
149  result = ff_dnn_get_input(&ctx->dnnctx, &model_input);
150  if (result != 0) {
151  av_log(ctx, AV_LOG_ERROR, "could not get input from the model\n");
152  return result;
153  }
154 
155  check = check_modelinput_inlink(&model_input, inlink);
156  if (check != 0) {
157  return check;
158  }
159 
160  return 0;
161 }
162 
164 {
166  av_assert0(desc);
167  return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components == 3;
168 }
169 
170 static int prepare_uv_scale(AVFilterLink *outlink)
171 {
172  AVFilterContext *context = outlink->src;
174  AVFilterLink *inlink = context->inputs[0];
175  enum AVPixelFormat fmt = inlink->format;
176 
177  if (isPlanarYUV(fmt)) {
178  if (inlink->w != outlink->w || inlink->h != outlink->h) {
179  if (fmt == AV_PIX_FMT_NV12) {
180  ctx->sws_uv_scale = sws_getContext(inlink->w >> 1, inlink->h >> 1, AV_PIX_FMT_YA8,
181  outlink->w >> 1, outlink->h >> 1, AV_PIX_FMT_YA8,
183  ctx->sws_uv_height = inlink->h >> 1;
184  } else {
186  int sws_src_h = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
187  int sws_src_w = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
188  int sws_dst_h = AV_CEIL_RSHIFT(outlink->h, desc->log2_chroma_h);
189  int sws_dst_w = AV_CEIL_RSHIFT(outlink->w, desc->log2_chroma_w);
190  ctx->sws_uv_scale = sws_getContext(sws_src_w, sws_src_h, AV_PIX_FMT_GRAY8,
191  sws_dst_w, sws_dst_h, AV_PIX_FMT_GRAY8,
193  ctx->sws_uv_height = sws_src_h;
194  }
195  }
196  }
197 
198  return 0;
199 }
200 
201 static int config_output(AVFilterLink *outlink)
202 {
203  AVFilterContext *context = outlink->src;
205  int result;
206  AVFilterLink *inlink = context->inputs[0];
207 
208  // have a try run in case that the dnn model resize the frame
209  result = ff_dnn_get_output(&ctx->dnnctx, inlink->w, inlink->h, &outlink->w, &outlink->h);
210  if (result != 0) {
211  av_log(ctx, AV_LOG_ERROR, "could not get output from the model\n");
212  return result;
213  }
214 
215  prepare_uv_scale(outlink);
216 
217  return 0;
218 }
219 
221 {
222  const AVPixFmtDescriptor *desc;
223  int uv_height;
224 
225  if (!ctx->sws_uv_scale) {
226  av_assert0(in->height == out->height && in->width == out->width);
228  uv_height = AV_CEIL_RSHIFT(in->height, desc->log2_chroma_h);
229  for (int i = 1; i < 3; ++i) {
230  int bytewidth = av_image_get_linesize(in->format, in->width, i);
231  if (bytewidth < 0) {
232  return AVERROR(EINVAL);
233  }
234  av_image_copy_plane(out->data[i], out->linesize[i],
235  in->data[i], in->linesize[i],
236  bytewidth, uv_height);
237  }
238  } else if (in->format == AV_PIX_FMT_NV12) {
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  } else {
242  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 1), in->linesize + 1,
243  0, ctx->sws_uv_height, out->data + 1, out->linesize + 1);
244  sws_scale(ctx->sws_uv_scale, (const uint8_t **)(in->data + 2), in->linesize + 2,
245  0, ctx->sws_uv_height, out->data + 2, out->linesize + 2);
246  }
247 
248  return 0;
249 }
250 
251 static int flush_frame(AVFilterLink *outlink, int64_t pts, int64_t *out_pts)
252 {
253  DnnProcessingContext *ctx = outlink->src->priv;
254  int ret;
255  DNNAsyncStatusType async_state;
256 
257  ret = ff_dnn_flush(&ctx->dnnctx);
258  if (ret != 0) {
259  return -1;
260  }
261 
262  do {
263  AVFrame *in_frame = NULL;
264  AVFrame *out_frame = NULL;
265  async_state = ff_dnn_get_result(&ctx->dnnctx, &in_frame, &out_frame);
266  if (out_frame) {
267  if (isPlanarYUV(in_frame->format))
268  copy_uv_planes(ctx, out_frame, in_frame);
269  av_frame_free(&in_frame);
270  ret = ff_filter_frame(outlink, out_frame);
271  if (ret < 0)
272  return ret;
273  if (out_pts)
274  *out_pts = out_frame->pts + pts;
275  }
276  av_usleep(5000);
277  } while (async_state >= DAST_NOT_READY);
278 
279  return 0;
280 }
281 
283 {
284  AVFilterLink *inlink = filter_ctx->inputs[0];
285  AVFilterLink *outlink = filter_ctx->outputs[0];
287  AVFrame *in = NULL, *out = NULL;
288  int64_t pts;
289  int ret, status;
290  int got_frame = 0;
291  int async_state;
292 
294 
295  do {
296  // drain all input frames
298  if (ret < 0)
299  return ret;
300  if (ret > 0) {
301  out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
302  if (!out) {
303  av_frame_free(&in);
304  return AVERROR(ENOMEM);
305  }
307  if (ff_dnn_execute_model(&ctx->dnnctx, in, out) != 0) {
308  return AVERROR(EIO);
309  }
310  }
311  } while (ret > 0);
312 
313  // drain all processed frames
314  do {
315  AVFrame *in_frame = NULL;
316  AVFrame *out_frame = NULL;
317  async_state = ff_dnn_get_result(&ctx->dnnctx, &in_frame, &out_frame);
318  if (out_frame) {
319  if (isPlanarYUV(in_frame->format))
320  copy_uv_planes(ctx, out_frame, in_frame);
321  av_frame_free(&in_frame);
322  ret = ff_filter_frame(outlink, out_frame);
323  if (ret < 0)
324  return ret;
325  got_frame = 1;
326  }
327  } while (async_state == DAST_SUCCESS);
328 
329  // if frame got, schedule to next filter
330  if (got_frame)
331  return 0;
332 
334  if (status == AVERROR_EOF) {
335  int64_t out_pts = pts;
336  ret = flush_frame(outlink, pts, &out_pts);
337  ff_outlink_set_status(outlink, status, out_pts);
338  return ret;
339  }
340  }
341 
343 
344  return 0;
345 }
346 
348 {
350 
351  sws_freeContext(context->sws_uv_scale);
352  ff_dnn_uninit(&context->dnnctx);
353 }
354 
356  {
357  .name = "default",
358  .type = AVMEDIA_TYPE_VIDEO,
359  .config_props = config_input,
360  },
361 };
362 
364  {
365  .name = "default",
366  .type = AVMEDIA_TYPE_VIDEO,
367  .config_props = config_output,
368  },
369 };
370 
372  .name = "dnn_processing",
373  .description = NULL_IF_CONFIG_SMALL("Apply DNN processing filter to the input."),
374  .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:347
AVFILTER_DNN_DEFINE_CLASS
AVFILTER_DNN_DEFINE_CLASS(dnn_processing)
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:55
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:1015
av_pix_fmt_desc_get
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:2965
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:374
pixdesc.h
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:486
DnnProcessingContext
Definition: vf_dnn_processing.c:37
AVFrame::width
int width
Definition: frame.h:446
AVOption
AVOption.
Definition: opt.h:357
AV_PIX_FMT_BGR24
@ AV_PIX_FMT_BGR24
packed RGB 8:8:8, 24bpp, BGRBGR...
Definition: pixfmt.h:76
preinit
static av_cold int preinit(AVFilterContext *ctx)
Definition: af_aresample.c:49
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:1206
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_dnn_processing.c:141
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:395
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:1442
dnn_get_width_idx_by_layout
static int dnn_get_width_idx_by_layout(DNNLayout layout)
Definition: dnn_interface.h:194
AVFilterContext::priv
void * priv
private data for use by the filter
Definition: avfilter.h:422
DnnContext
Definition: dnn_interface.h:141
filter_ctx
static FilteringContext * filter_ctx
Definition: transcode.c:52
ff_dnn_filter_init_child_class
int ff_dnn_filter_init_child_class(AVFilterContext *filter)
Definition: dnn_filter_common.c:61
pts
static int64_t pts
Definition: transcode_aac.c:644
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:251
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:220
AV_CEIL_RSHIFT
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:59
pix_fmt
static enum AVPixelFormat pix_fmt
Definition: demux_decode.c:41
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: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:198
ctx
AVFormatContext * ctx
Definition: movenc.c:49
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:84
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:158
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
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:62
time.h
dnn_processing_inputs
static const AVFilterPad dnn_processing_inputs[]
Definition: vf_dnn_processing.c:355
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:203
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:1389
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:94
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:2102
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:461
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:371
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:256
activate
static int activate(AVFilterContext *filter_ctx)
Definition: vf_dnn_processing.c:282
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:446
status
ov_status_e status
Definition: dnn_backend_openvino.c:101
sws_freeContext
void sws_freeContext(struct SwsContext *swsContext)
Free the swscaler context swsContext.
Definition: utils.c:2433
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:163
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:72
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:245
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:163
LOG_FORMAT_CHANNEL_MISMATCH
#define LOG_FORMAT_CHANNEL_MISMATCH()
Definition: vf_dnn_processing.c:77
DNNData::dims
int dims[4]
Definition: dnn_interface.h:67
desc
const char * desc
Definition: libsvtav1.c:79
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:199
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:204
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:73
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:170
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:419
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:208
ff_dnn_execute_model
int ff_dnn_execute_model(DnnContext *ctx, AVFrame *in_frame, AVFrame *out_frame)
Definition: dnn_filter_common.c:171
DAST_NOT_READY
@ DAST_NOT_READY
Definition: dnn_interface.h:48
SwsContext
Definition: swscale_internal.h:301
DNNAsyncStatusType
DNNAsyncStatusType
Definition: dnn_interface.h:45
AV_OPT_TYPE_CONST
@ AV_OPT_TYPE_CONST
Definition: opt.h:254
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:363
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:2885
config_output
static int config_output(AVFilterLink *outlink)
Definition: vf_dnn_processing.c:201