git.sesse.net Git - ffmpeg/blob - libavfilter/vf_libvmaf.c

   1 /*
   2  * Copyright (c) 2017 Ronald S. Bultje <rsbultje@gmail.com>
   3  * Copyright (c) 2017 Ashish Pratap Singh <ashk43712@gmail.com>
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file
  24  * Calculate the VMAF between two input videos.
  25  */
  26
  27 #include <pthread.h>
  28 #include <libvmaf.h>
  29 #include "libavutil/avstring.h"
  30 #include "libavutil/opt.h"
  31 #include "libavutil/pixdesc.h"
  32 #include "avfilter.h"
  33 #include "drawutils.h"
  34 #include "formats.h"
  35 #include "framesync.h"
  36 #include "internal.h"
  37 #include "video.h"
  38
  39 typedef struct LIBVMAFContext {
  40     const AVClass *class;
  41     FFFrameSync fs;
  42     const AVPixFmtDescriptor *desc;
  43     int width;
  44     int height;
  45     double vmaf_score;
  46     int vmaf_thread_created;
  47     pthread_t vmaf_thread;
  48     pthread_mutex_t lock;
  49     pthread_cond_t cond;
  50     int eof;
  51     AVFrame *gmain;
  52     AVFrame *gref;
  53     int frame_set;
  54     char *model_path;
  55     char *log_path;
  56     char *log_fmt;
  57     int disable_clip;
  58     int disable_avx;
  59     int enable_transform;
  60     int phone_model;
  61     int psnr;
  62     int ssim;
  63     int ms_ssim;
  64     char *pool;
  65     int n_threads;
  66     int n_subsample;
  67     int enable_conf_interval;
  68     int error;
  69 } LIBVMAFContext;
  70
  71 #define OFFSET(x) offsetof(LIBVMAFContext, x)
  72 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
  73
  74 static const AVOption libvmaf_options[] = {
  75     {"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},
  76     {"log_path",  "Set the file path to be used to store logs.",                        OFFSET(log_path), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 1, FLAGS},
  77     {"log_fmt",  "Set the format of the log (xml or json).",                            OFFSET(log_fmt), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 1, FLAGS},
  78     {"enable_transform",  "Enables transform for computing vmaf.",                      OFFSET(enable_transform), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
  79     {"phone_model",  "Invokes the phone model that will generate higher VMAF scores.",  OFFSET(phone_model), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
  80     {"psnr",  "Enables computing psnr along with vmaf.",                                OFFSET(psnr), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
  81     {"ssim",  "Enables computing ssim along with vmaf.",                                OFFSET(ssim), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
  82     {"ms_ssim",  "Enables computing ms-ssim along with vmaf.",                          OFFSET(ms_ssim), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
  83     {"pool",  "Set the pool method to be used for computing vmaf.",                     OFFSET(pool), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 1, FLAGS},
  84     {"n_threads", "Set number of threads to be used when computing vmaf.",              OFFSET(n_threads), AV_OPT_TYPE_INT, {.i64=0}, 0, UINT_MAX, FLAGS},
  85     {"n_subsample", "Set interval for frame subsampling used when computing vmaf.",     OFFSET(n_subsample), AV_OPT_TYPE_INT, {.i64=1}, 1, UINT_MAX, FLAGS},
  86     {"enable_conf_interval",  "Enables confidence interval.",                           OFFSET(enable_conf_interval), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS},
  87     { NULL }
  88 };
  89
  90 FRAMESYNC_DEFINE_CLASS(libvmaf, LIBVMAFContext, fs);
  91
  92 #define read_frame_fn(type, bits)                                               \
  93     static int read_frame_##bits##bit(float *ref_data, float *main_data,        \
  94                                       float *temp_data, int stride, void *ctx)  \
  95 {                                                                               \
  96     LIBVMAFContext *s = (LIBVMAFContext *) ctx;                                 \
  97     int ret;                                                                    \
  98     \
  99     pthread_mutex_lock(&s->lock);                                               \
 100     \
 101     while (!s->frame_set && !s->eof) {                                          \
 102         pthread_cond_wait(&s->cond, &s->lock);                                  \
 103     }                                                                           \
 104     \
 105     if (s->frame_set) {                                                         \
 106         int ref_stride = s->gref->linesize[0];                                  \
 107         int main_stride = s->gmain->linesize[0];                                \
 108         \
 109         const type *ref_ptr = (const type *) s->gref->data[0];                  \
 110         const type *main_ptr = (const type *) s->gmain->data[0];                \
 111         \
 112         float *ptr = ref_data;                                                  \
 113         \
 114         int h = s->height;                                                      \
 115         int w = s->width;                                                       \
 116         \
 117         int i,j;                                                                \
 118         \
 119         for (i = 0; i < h; i++) {                                               \
 120             for ( j = 0; j < w; j++) {                                          \
 121                 ptr[j] = (float)ref_ptr[j];                                     \
 122             }                                                                   \
 123             ref_ptr += ref_stride / sizeof(*ref_ptr);                           \
 124             ptr += stride / sizeof(*ptr);                                       \
 125         }                                                                       \
 126         \
 127         ptr = main_data;                                                        \
 128         \
 129         for (i = 0; i < h; i++) {                                               \
 130             for (j = 0; j < w; j++) {                                           \
 131                 ptr[j] = (float)main_ptr[j];                                    \
 132             }                                                                   \
 133             main_ptr += main_stride / sizeof(*main_ptr);                        \
 134             ptr += stride / sizeof(*ptr);                                       \
 135         }                                                                       \
 136     }                                                                           \
 137     \
 138     ret = !s->frame_set;                                                        \
 139     \
 140     av_frame_unref(s->gref);                                                    \
 141     av_frame_unref(s->gmain);                                                   \
 142     s->frame_set = 0;                                                           \
 143     \
 144     pthread_cond_signal(&s->cond);                                              \
 145     pthread_mutex_unlock(&s->lock);                                             \
 146     \
 147     if (ret) {                                                                  \
 148         return 2;                                                               \
 149     }                                                                           \
 150     \
 151     return 0;                                                                   \
 152 }
 153
 154 read_frame_fn(uint8_t, 8);
 155 read_frame_fn(uint16_t, 10);
 156
 157 static void compute_vmaf_score(LIBVMAFContext *s)
 158 {
 159     int (*read_frame)(float *ref_data, float *main_data, float *temp_data,
 160                       int stride, void *ctx);
 161     char *format;
 162
 163     if (s->desc->comp[0].depth <= 8) {
 164         read_frame = read_frame_8bit;
 165     } else {
 166         read_frame = read_frame_10bit;
 167     }
 168
 169     format = (char *) s->desc->name;
 170
 171     s->error = compute_vmaf(&s->vmaf_score, format, s->width, s->height,
 172                             read_frame, s, s->model_path, s->log_path,
 173                             s->log_fmt, 0, 0, s->enable_transform,
 174                             s->phone_model, s->psnr, s->ssim,
 175                             s->ms_ssim, s->pool,
 176                             s->n_threads, s->n_subsample, s->enable_conf_interval);
 177 }
 178
 179 static void *call_vmaf(void *ctx)
 180 {
 181     LIBVMAFContext *s = (LIBVMAFContext *) ctx;
 182     compute_vmaf_score(s);
 183     if (!s->error) {
 184         av_log(ctx, AV_LOG_INFO, "VMAF score: %f\n",s->vmaf_score);
 185     } else {
 186         pthread_mutex_lock(&s->lock);
 187         pthread_cond_signal(&s->cond);
 188         pthread_mutex_unlock(&s->lock);
 189     }
 190     pthread_exit(NULL);
 191     return NULL;
 192 }
 193
 194 static int do_vmaf(FFFrameSync *fs)
 195 {
 196     AVFilterContext *ctx = fs->parent;
 197     LIBVMAFContext *s = ctx->priv;
 198     AVFrame *master, *ref;
 199     int ret;
 200
 201     ret = ff_framesync_dualinput_get(fs, &master, &ref);
 202     if (ret < 0)
 203         return ret;
 204     if (!ref)
 205         return ff_filter_frame(ctx->outputs[0], master);
 206
 207     pthread_mutex_lock(&s->lock);
 208
 209     while (s->frame_set && !s->error) {
 210         pthread_cond_wait(&s->cond, &s->lock);
 211     }
 212
 213     if (s->error) {
 214         av_log(ctx, AV_LOG_ERROR,
 215                "libvmaf encountered an error, check log for details\n");
 216         pthread_mutex_unlock(&s->lock);
 217         return AVERROR(EINVAL);
 218     }
 219
 220     av_frame_ref(s->gref, ref);
 221     av_frame_ref(s->gmain, master);
 222
 223     s->frame_set = 1;
 224
 225     pthread_cond_signal(&s->cond);
 226     pthread_mutex_unlock(&s->lock);
 227
 228     return ff_filter_frame(ctx->outputs[0], master);
 229 }
 230
 231 static av_cold int init(AVFilterContext *ctx)
 232 {
 233     LIBVMAFContext *s = ctx->priv;
 234
 235     s->gref = av_frame_alloc();
 236     s->gmain = av_frame_alloc();
 237     s->error = 0;
 238
 239     s->vmaf_thread_created = 0;
 240     pthread_mutex_init(&s->lock, NULL);
 241     pthread_cond_init (&s->cond, NULL);
 242
 243     s->fs.on_event = do_vmaf;
 244     return 0;
 245 }
 246
 247 static int query_formats(AVFilterContext *ctx)
 248 {
 249     static const enum AVPixelFormat pix_fmts[] = {
 250         AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV420P,
 251         AV_PIX_FMT_YUV444P10LE, AV_PIX_FMT_YUV422P10LE, AV_PIX_FMT_YUV420P10LE,
 252         AV_PIX_FMT_NONE
 253     };
 254
 255     AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
 256     if (!fmts_list)
 257         return AVERROR(ENOMEM);
 258     return ff_set_common_formats(ctx, fmts_list);
 259 }
 260
 261
 262 static int config_input_ref(AVFilterLink *inlink)
 263 {
 264     AVFilterContext *ctx  = inlink->dst;
 265     LIBVMAFContext *s = ctx->priv;
 266     int th;
 267
 268     if (ctx->inputs[0]->w != ctx->inputs[1]->w ||
 269         ctx->inputs[0]->h != ctx->inputs[1]->h) {
 270         av_log(ctx, AV_LOG_ERROR, "Width and height of input videos must be same.\n");
 271         return AVERROR(EINVAL);
 272     }
 273     if (ctx->inputs[0]->format != ctx->inputs[1]->format) {
 274         av_log(ctx, AV_LOG_ERROR, "Inputs must be of same pixel format.\n");
 275         return AVERROR(EINVAL);
 276     }
 277
 278     s->desc = av_pix_fmt_desc_get(inlink->format);
 279     s->width = ctx->inputs[0]->w;
 280     s->height = ctx->inputs[0]->h;
 281
 282     th = pthread_create(&s->vmaf_thread, NULL, call_vmaf, (void *) s);
 283     if (th) {
 284         av_log(ctx, AV_LOG_ERROR, "Thread creation failed.\n");
 285         return AVERROR(EINVAL);
 286     }
 287     s->vmaf_thread_created = 1;
 288
 289     return 0;
 290 }
 291
 292 static int config_output(AVFilterLink *outlink)
 293 {
 294     AVFilterContext *ctx = outlink->src;
 295     LIBVMAFContext *s = ctx->priv;
 296     AVFilterLink *mainlink = ctx->inputs[0];
 297     int ret;
 298
 299     ret = ff_framesync_init_dualinput(&s->fs, ctx);
 300     if (ret < 0)
 301         return ret;
 302     outlink->w = mainlink->w;
 303     outlink->h = mainlink->h;
 304     outlink->time_base = mainlink->time_base;
 305     outlink->sample_aspect_ratio = mainlink->sample_aspect_ratio;
 306     outlink->frame_rate = mainlink->frame_rate;
 307     if ((ret = ff_framesync_configure(&s->fs)) < 0)
 308         return ret;
 309
 310     return 0;
 311 }
 312
 313 static int activate(AVFilterContext *ctx)
 314 {
 315     LIBVMAFContext *s = ctx->priv;
 316     return ff_framesync_activate(&s->fs);
 317 }
 318
 319 static av_cold void uninit(AVFilterContext *ctx)
 320 {
 321     LIBVMAFContext *s = ctx->priv;
 322
 323     ff_framesync_uninit(&s->fs);
 324
 325     pthread_mutex_lock(&s->lock);
 326     s->eof = 1;
 327     pthread_cond_signal(&s->cond);
 328     pthread_mutex_unlock(&s->lock);
 329
 330     if (s->vmaf_thread_created)
 331     {
 332         pthread_join(s->vmaf_thread, NULL);
 333         s->vmaf_thread_created = 0;
 334     }
 335
 336     av_frame_free(&s->gref);
 337     av_frame_free(&s->gmain);
 338
 339     pthread_mutex_destroy(&s->lock);
 340     pthread_cond_destroy(&s->cond);
 341 }
 342
 343 static const AVFilterPad libvmaf_inputs[] = {
 344     {
 345         .name         = "main",
 346         .type         = AVMEDIA_TYPE_VIDEO,
 347     },{
 348         .name         = "reference",
 349         .type         = AVMEDIA_TYPE_VIDEO,
 350         .config_props = config_input_ref,
 351     },
 352     { NULL }
 353 };
 354
 355 static const AVFilterPad libvmaf_outputs[] = {
 356     {
 357         .name          = "default",
 358         .type          = AVMEDIA_TYPE_VIDEO,
 359         .config_props  = config_output,
 360     },
 361     { NULL }
 362 };
 363
 364 AVFilter ff_vf_libvmaf = {
 365     .name          = "libvmaf",
 366     .description   = NULL_IF_CONFIG_SMALL("Calculate the VMAF between two video streams."),
 367     .preinit       = libvmaf_framesync_preinit,
 368     .init          = init,
 369     .uninit        = uninit,
 370     .query_formats = query_formats,
 371     .activate      = activate,
 372     .priv_size     = sizeof(LIBVMAFContext),
 373     .priv_class    = &libvmaf_class,
 374     .inputs        = libvmaf_inputs,
 375     .outputs       = libvmaf_outputs,
 376 };