1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
|
/*
* Copyright (c) 2017 Ronald S. Bultje <rsbultje@gmail.com>
* Copyright (c) 2017 Ashish Pratap Singh <ashk43712@gmail.com>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* Calculate the VMAF between two input videos.
*/
#include <inttypes.h>
#include <pthread.h>
#include <libvmaf.h>
#include "libavutil/avstring.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "avfilter.h"
#include "dualinput.h"
#include "drawutils.h"
#include "formats.h"
#include "internal.h"
#include "video.h"
typedef struct LIBVMAFContext {
const AVClass *class;
FFDualInputContext dinput;
const AVPixFmtDescriptor *desc;
char *format;
int width;
int height;
double vmaf_score;
pthread_t vmaf_thread;
pthread_mutex_t lock;
pthread_cond_t cond;
int eof;
AVFrame *gmain;
AVFrame *gref;
int frame_set;
char *model_path;
char *log_path;
char *log_fmt;
int disable_clip;
int disable_avx;
int enable_transform;
int phone_model;
int psnr;
int ssim;
int ms_ssim;
char *pool;
} LIBVMAFContext;
#define OFFSET(x) offsetof(LIBVMAFContext, x)
#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
static const AVOption libvmaf_options[] = {
{"model_path", "Set the model to be used for computing vmaf.", OFFSET(model_path), AV_OPT_TYPE_STRING, {.str="/usr/local/share/model/vmaf_v0.6.1.pkl"}, 0, 1, FLAGS},
{"log_path", "Set the file path to be used to store logs.", OFFSET(log_path), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 1, FLAGS},
{"log_fmt", "Set the format of the log (xml or json).", OFFSET(log_fmt), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 1, FLAGS},
{"enable_transform", "Enables transform for computing vmaf.", OFFSET(enable_transform), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
{"phone_model", "Invokes the phone model that will generate higher VMAF scores.", OFFSET(phone_model), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
{"psnr", "Enables computing psnr along with vmaf.", OFFSET(psnr), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
{"ssim", "Enables computing ssim along with vmaf.", OFFSET(ssim), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
{"ms_ssim", "Enables computing ms-ssim along with vmaf.", OFFSET(ms_ssim), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
{"pool", "Set the pool method to be used for computing vmaf.", OFFSET(pool), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 1, FLAGS},
{ NULL }
};
AVFILTER_DEFINE_CLASS(libvmaf);
#define read_frame_fn(type, bits) \
static int read_frame_##bits##bit(float *ref_data, float *main_data, \
float *temp_data, int stride, \
double *score, void *ctx) \
{ \
LIBVMAFContext *s = (LIBVMAFContext *) ctx; \
int ret; \
\
pthread_mutex_lock(&s->lock); \
\
while (!s->frame_set && !s->eof) { \
pthread_cond_wait(&s->cond, &s->lock); \
} \
\
if (s->frame_set) { \
int ref_stride = s->gref->linesize[0]; \
int main_stride = s->gmain->linesize[0]; \
\
const type *ref_ptr = (const type *) s->gref->data[0]; \
const type *main_ptr = (const type *) s->gmain->data[0]; \
\
float *ptr = ref_data; \
\
int h = s->height; \
int w = s->width; \
\
int i,j; \
\
for (i = 0; i < h; i++) { \
for ( j = 0; j < w; j++) { \
ptr[j] = (float)ref_ptr[j]; \
} \
ref_ptr += ref_stride / sizeof(*ref_ptr); \
ptr += stride / sizeof(*ptr); \
} \
\
ptr = main_data; \
\
for (i = 0; i < h; i++) { \
for (j = 0; j < w; j++) { \
ptr[j] = (float)main_ptr[j]; \
} \
main_ptr += main_stride / sizeof(*main_ptr); \
ptr += stride / sizeof(*ptr); \
} \
} \
\
ret = !s->frame_set; \
\
s->frame_set = 0; \
\
pthread_cond_signal(&s->cond); \
pthread_mutex_unlock(&s->lock); \
\
if (ret) { \
return 2; \
} \
\
return 0; \
}
read_frame_fn(uint8_t, 8);
read_frame_fn(uint16_t, 10);
static void compute_vmaf_score(LIBVMAFContext *s)
{
int (*read_frame)(float *ref_data, float *main_data, float *temp_data,
int stride, double *score, void *ctx);
if (s->desc->comp[0].depth <= 8) {
read_frame = read_frame_8bit;
} else {
read_frame = read_frame_10bit;
}
s->vmaf_score = compute_vmaf(s->format, s->width, s->height, read_frame, s,
s->model_path, s->log_path, s->log_fmt, 0, 0,
s->enable_transform, s->phone_model, s->psnr,
s->ssim, s->ms_ssim, s->pool);
}
static void *call_vmaf(void *ctx)
{
LIBVMAFContext *s = (LIBVMAFContext *) ctx;
compute_vmaf_score(s);
av_log(ctx, AV_LOG_INFO, "VMAF score: %f\n",s->vmaf_score);
pthread_exit(NULL);
}
static AVFrame *do_vmaf(AVFilterContext *ctx, AVFrame *main, const AVFrame *ref)
{
LIBVMAFContext *s = ctx->priv;
pthread_mutex_lock(&s->lock);
while (s->frame_set != 0) {
pthread_cond_wait(&s->cond, &s->lock);
}
av_frame_ref(s->gref, ref);
av_frame_ref(s->gmain, main);
s->frame_set = 1;
pthread_cond_signal(&s->cond);
pthread_mutex_unlock(&s->lock);
return main;
}
static av_cold int init(AVFilterContext *ctx)
{
LIBVMAFContext *s = ctx->priv;
s->gref = av_frame_alloc();
s->gmain = av_frame_alloc();
pthread_mutex_init(&s->lock, NULL);
pthread_cond_init (&s->cond, NULL);
s->dinput.process = do_vmaf;
return 0;
}
static int query_formats(AVFilterContext *ctx)
{
static const enum AVPixelFormat pix_fmts[] = {
AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV420P,
AV_PIX_FMT_YUV444P10LE, AV_PIX_FMT_YUV422P10LE, AV_PIX_FMT_YUV420P10LE,
AV_PIX_FMT_NONE
};
AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
if (!fmts_list)
return AVERROR(ENOMEM);
return ff_set_common_formats(ctx, fmts_list);
}
static int config_input_ref(AVFilterLink *inlink)
{
AVFilterContext *ctx = inlink->dst;
LIBVMAFContext *s = ctx->priv;
int th;
if (ctx->inputs[0]->w != ctx->inputs[1]->w ||
ctx->inputs[0]->h != ctx->inputs[1]->h) {
av_log(ctx, AV_LOG_ERROR, "Width and height of input videos must be same.\n");
return AVERROR(EINVAL);
}
if (ctx->inputs[0]->format != ctx->inputs[1]->format) {
av_log(ctx, AV_LOG_ERROR, "Inputs must be of same pixel format.\n");
return AVERROR(EINVAL);
}
s->desc = av_pix_fmt_desc_get(inlink->format);
s->width = ctx->inputs[0]->w;
s->height = ctx->inputs[0]->h;
th = pthread_create(&s->vmaf_thread, NULL, call_vmaf, (void *) s);
if (th) {
av_log(ctx, AV_LOG_ERROR, "Thread creation failed.\n");
return AVERROR(EINVAL);
}
return 0;
}
static int config_output(AVFilterLink *outlink)
{
AVFilterContext *ctx = outlink->src;
LIBVMAFContext *s = ctx->priv;
AVFilterLink *mainlink = ctx->inputs[0];
int ret;
outlink->w = mainlink->w;
outlink->h = mainlink->h;
outlink->time_base = mainlink->time_base;
outlink->sample_aspect_ratio = mainlink->sample_aspect_ratio;
outlink->frame_rate = mainlink->frame_rate;
if ((ret = ff_dualinput_init(ctx, &s->dinput)) < 0)
return ret;
return 0;
}
static int filter_frame(AVFilterLink *inlink, AVFrame *inpicref)
{
LIBVMAFContext *s = inlink->dst->priv;
return ff_dualinput_filter_frame(&s->dinput, inlink, inpicref);
}
static int request_frame(AVFilterLink *outlink)
{
LIBVMAFContext *s = outlink->src->priv;
return ff_dualinput_request_frame(&s->dinput, outlink);
}
static av_cold void uninit(AVFilterContext *ctx)
{
LIBVMAFContext *s = ctx->priv;
ff_dualinput_uninit(&s->dinput);
pthread_mutex_lock(&s->lock);
s->eof = 1;
pthread_cond_signal(&s->cond);
pthread_mutex_unlock(&s->lock);
pthread_join(s->vmaf_thread, NULL);
av_frame_free(&s->gref);
av_frame_free(&s->gmain);
pthread_mutex_destroy(&s->lock);
pthread_cond_destroy(&s->cond);
}
static const AVFilterPad libvmaf_inputs[] = {
{
.name = "main",
.type = AVMEDIA_TYPE_VIDEO,
.filter_frame = filter_frame,
},{
.name = "reference",
.type = AVMEDIA_TYPE_VIDEO,
.filter_frame = filter_frame,
.config_props = config_input_ref,
},
{ NULL }
};
static const AVFilterPad libvmaf_outputs[] = {
{
.name = "default",
.type = AVMEDIA_TYPE_VIDEO,
.config_props = config_output,
.request_frame = request_frame,
},
{ NULL }
};
AVFilter ff_vf_libvmaf = {
.name = "libvmaf",
.description = NULL_IF_CONFIG_SMALL("Calculate the VMAF between two video streams."),
.init = init,
.uninit = uninit,
.query_formats = query_formats,
.priv_size = sizeof(LIBVMAFContext),
.priv_class = &libvmaf_class,
.inputs = libvmaf_inputs,
.outputs = libvmaf_outputs,
};
|