int main(int argc, char **argv) {
VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = {NULL};
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_cnt = 0;
vpx_image_t raw;
vpx_codec_err_t res;
unsigned int width;
unsigned int height;
int speed;
int frame_avail;
int got_data;
int flags = 0;
unsigned int i;
int pts = 0; // PTS starts at 0.
int frame_duration = 1; // 1 timebase tick per frame.
int layering_mode = 0;
int layer_flags[VPX_TS_MAX_PERIODICITY] = {0};
int flag_periodicity = 1;
#if VPX_ENCODER_ABI_VERSION > (4 + VPX_CODEC_ABI_VERSION)
vpx_svc_layer_id_t layer_id = {0, 0};
#else
vpx_svc_layer_id_t layer_id = {0};
#endif
const VpxInterface *encoder = NULL;
FILE *infile = NULL;
struct RateControlMetrics rc;
int64_t cx_time = 0;
const int min_args_base = 11;
#if CONFIG_VP9_HIGHBITDEPTH
vpx_bit_depth_t bit_depth = VPX_BITS_8;
int input_bit_depth = 8;
const int min_args = min_args_base + 1;
#else
const int min_args = min_args_base;
#endif // CONFIG_VP9_HIGHBITDEPTH
double sum_bitrate = 0.0;
double sum_bitrate2 = 0.0;
double framerate = 30.0;
exec_name = argv[0];
// Check usage and arguments.
if (argc < min_args) {
#if CONFIG_VP9_HIGHBITDEPTH
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n", argv[0]);
#else
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> \n", argv[0]);
#endif // CONFIG_VP9_HIGHBITDEPTH
}
encoder = get_vpx_encoder_by_name(argv[3]);
if (!encoder)
die("Unsupported codec.");
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
width = strtol(argv[4], NULL, 0);
height = strtol(argv[5], NULL, 0);
if (width < 16 || width % 2 || height < 16 || height % 2) {
die("Invalid resolution: %d x %d", width, height);
}
layering_mode = strtol(argv[10], NULL, 0);
if (layering_mode < 0 || layering_mode > 12) {
die("Invalid layering mode (0..12) %s", argv[10]);
}
if (argc != min_args + mode_to_num_layers[layering_mode]) {
die("Invalid number of arguments");
}
#if CONFIG_VP9_HIGHBITDEPTH
switch (strtol(argv[argc-1], NULL, 0)) {
case 8:
bit_depth = VPX_BITS_8;
input_bit_depth = 8;
break;
case 10:
bit_depth = VPX_BITS_10;
input_bit_depth = 10;
break;
case 12:
bit_depth = VPX_BITS_12;
input_bit_depth = 12;
break;
default:
die("Invalid bit depth (8, 10, 12) %s", argv[argc-1]);
}
if (!vpx_img_alloc(&raw,
bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 :
VPX_IMG_FMT_I42016,
width, height, 32)) {
die("Failed to allocate image", width, height);
}
#else
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
die("Failed to allocate image", width, height);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Populate encoder configuration.
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) {
printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
return EXIT_FAILURE;
}
// Update the default configuration with our settings.
cfg.g_w = width;
cfg.g_h = height;
#if CONFIG_VP9_HIGHBITDEPTH
if (bit_depth != VPX_BITS_8) {
cfg.g_bit_depth = bit_depth;
cfg.g_input_bit_depth = input_bit_depth;
cfg.g_profile = 2;
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Timebase format e.g. 30fps: numerator=1, demoninator = 30.
cfg.g_timebase.num = strtol(argv[6], NULL, 0);
cfg.g_timebase.den = strtol(argv[7], NULL, 0);
speed = strtol(argv[8], NULL, 0);
if (speed < 0) {
die("Invalid speed setting: must be positive");
}
for (i = min_args_base;
(int)i < min_args_base + mode_to_num_layers[layering_mode];
++i) {
rc.layer_target_bitrate[i - 11] = strtol(argv[i], NULL, 0);
if (strncmp(encoder->name, "vp8", 3) == 0)
cfg.ts_target_bitrate[i - 11] = rc.layer_target_bitrate[i - 11];
else if (strncmp(encoder->name, "vp9", 3) == 0)
cfg.layer_target_bitrate[i - 11] = rc.layer_target_bitrate[i - 11];
}
// Real time parameters.
cfg.rc_dropframe_thresh = strtol(argv[9], NULL, 0);
cfg.rc_end_usage = VPX_CBR;
cfg.rc_min_quantizer = 2;
cfg.rc_max_quantizer = 56;
if (strncmp(encoder->name, "vp9", 3) == 0)
cfg.rc_max_quantizer = 52;
cfg.rc_undershoot_pct = 50;
cfg.rc_overshoot_pct = 50;
cfg.rc_buf_initial_sz = 500;
cfg.rc_buf_optimal_sz = 600;
cfg.rc_buf_sz = 1000;
// Disable dynamic resizing by default.
cfg.rc_resize_allowed = 0;
// Use 1 thread as default.
cfg.g_threads = 1;
// Enable error resilient mode.
cfg.g_error_resilient = 1;
cfg.g_lag_in_frames = 0;
cfg.kf_mode = VPX_KF_AUTO;
// Disable automatic keyframe placement.
cfg.kf_min_dist = cfg.kf_max_dist = 3000;
cfg.temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
set_temporal_layer_pattern(layering_mode,
&cfg,
layer_flags,
&flag_periodicity);
set_rate_control_metrics(&rc, &cfg);
// Target bandwidth for the whole stream.
// Set to layer_target_bitrate for highest layer (total bitrate).
cfg.rc_target_bitrate = rc.layer_target_bitrate[cfg.ts_number_layers - 1];
// Open input file.
if (!(infile = fopen(argv[1], "rb"))) {
die("Failed to open %s for reading", argv[1]);
}
framerate = cfg.g_timebase.den / cfg.g_timebase.num;
// Open an output file for each stream.
for (i = 0; i < cfg.ts_number_layers; ++i) {
char file_name[PATH_MAX];
VpxVideoInfo info;
info.codec_fourcc = encoder->fourcc;
info.frame_width = cfg.g_w;
info.frame_height = cfg.g_h;
info.time_base.numerator = cfg.g_timebase.num;
info.time_base.denominator = cfg.g_timebase.den;
snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
if (!outfile[i])
die("Failed to open %s for writing", file_name);
assert(outfile[i] != NULL);
}
// No spatial layers in this encoder.
cfg.ss_number_layers = 1;
// Initialize codec.
#if CONFIG_VP9_HIGHBITDEPTH
if (vpx_codec_enc_init(
&codec, encoder->codec_interface(), &cfg,
bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
#else
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
#endif // CONFIG_VP9_HIGHBITDEPTH
die_codec(&codec, "Failed to initialize encoder");
// configuration for vp8/vp9
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kDenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 0);
} else if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_svc_extra_cfg_t svc_params;
vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0);
vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, 0);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 0);
vpx_codec_control(&codec, VP9E_SET_TUNE_CONTENT, 0);
vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (cfg.g_threads >> 1));
if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1: 0))
die_codec(&codec, "Failed to set SVC");
for (i = 0; i < cfg.ts_number_layers; ++i) {
svc_params.max_quantizers[i] = cfg.rc_max_quantizer;
svc_params.min_quantizers[i] = cfg.rc_min_quantizer;
}
svc_params.scaling_factor_num[0] = cfg.g_h;
svc_params.scaling_factor_den[0] = cfg.g_h;
vpx_codec_control(&codec, VP9E_SET_SVC_PARAMETERS, &svc_params);
}
int main(int argc, char **argv) {
FILE *infile = NULL;
int w, h;
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
vpx_image_t raw;
vpx_codec_err_t res;
vpx_fixed_buf_t stats;
const VpxInterface *encoder = NULL;
const int fps = 30; // TODO(dkovalev) add command line argument
const int bitrate = 200; // kbit/s TODO(dkovalev) add command line argument
const char *const codec_arg = argv[1];
const char *const width_arg = argv[2];
const char *const height_arg = argv[3];
const char *const infile_arg = argv[4];
const char *const outfile_arg = argv[5];
exec_name = argv[0];
if (argc != 6) die("Invalid number of arguments.");
encoder = get_vpx_encoder_by_name(codec_arg);
if (!encoder) die("Unsupported codec.");
w = strtol(width_arg, NULL, 0);
h = strtol(height_arg, NULL, 0);
if (w <= 0 || h <= 0 || (w % 2) != 0 || (h % 2) != 0)
die("Invalid frame size: %dx%d", w, h);
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, w, h, 1))
die("Failed to allocate image", w, h);
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
// Configuration
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) die_codec(&codec, "Failed to get default codec config.");
cfg.g_w = w;
cfg.g_h = h;
cfg.g_timebase.num = 1;
cfg.g_timebase.den = fps;
cfg.rc_target_bitrate = bitrate;
if (!(infile = fopen(infile_arg, "rb")))
die("Failed to open %s for reading", infile_arg);
// Pass 0
cfg.g_pass = VPX_RC_FIRST_PASS;
stats = pass0(&raw, infile, encoder, &cfg);
// Pass 1
rewind(infile);
cfg.g_pass = VPX_RC_LAST_PASS;
cfg.rc_twopass_stats_in = stats;
pass1(&raw, infile, outfile_arg, encoder, &cfg);
free(stats.buf);
vpx_img_free(&raw);
fclose(infile);
return EXIT_SUCCESS;
}
int main(int argc, char **argv) {
FILE *infile = NULL;
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_count = 0;
vpx_image_t raw;
vpx_codec_err_t res;
VpxVideoInfo info;
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
const int fps = 2; // TODO(dkovalev) add command line argument
const double bits_per_pixel_per_frame = 0.067;
exec_name = argv[0];
if (argc != 6)
die("Invalid number of arguments");
memset(&info, 0, sizeof(info));
encoder = get_vpx_encoder_by_name(argv[1]);
if (encoder == NULL) {
die("Unsupported codec.");
}
assert(encoder != NULL);
info.codec_fourcc = encoder->fourcc;
info.frame_width = strtol(argv[2], NULL, 0);
info.frame_height = strtol(argv[3], NULL, 0);
info.time_base.numerator = 1;
info.time_base.denominator = fps;
if (info.frame_width <= 0 ||
info.frame_height <= 0 ||
(info.frame_width % 2) != 0 ||
(info.frame_height % 2) != 0) {
die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
}
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
info.frame_height, 1)) {
die("Failed to allocate image.");
}
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res)
die_codec(&codec, "Failed to get default codec config.");
cfg.g_w = info.frame_width;
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
cfg.rc_target_bitrate = (unsigned int)(bits_per_pixel_per_frame * cfg.g_w *
cfg.g_h * fps / 1000);
cfg.g_lag_in_frames = 0;
writer = vpx_video_writer_open(argv[5], kContainerIVF, &info);
if (!writer)
die("Failed to open %s for writing.", argv[5]);
if (!(infile = fopen(argv[4], "rb")))
die("Failed to open %s for reading.", argv[4]);
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
die_codec(&codec, "Failed to initialize encoder");
// Encode frames.
while (vpx_img_read(&raw, infile)) {
++frame_count;
if (frame_count == 22 && encoder->fourcc == VP8_FOURCC) {
set_roi_map(&cfg, &codec);
} else if (frame_count == 33) {
set_active_map(&cfg, &codec);
} else if (frame_count == 44) {
unset_active_map(&cfg, &codec);
}
encode_frame(&codec, &raw, frame_count, writer);
}
// Flush encoder.
while (encode_frame(&codec, NULL, -1, writer)) {}
printf("\n");
fclose(infile);
printf("Processed %d frames.\n", frame_count);
vpx_img_free(&raw);
if (vpx_codec_destroy(&codec))
die_codec(&codec, "Failed to destroy codec.");
vpx_video_writer_close(writer);
return EXIT_SUCCESS;
}
int main(int argc, char **argv) {
FILE *infile = NULL;
// Encoder
vpx_codec_ctx_t ecodec = {0};
vpx_codec_enc_cfg_t cfg = {0};
unsigned int frame_in = 0;
vpx_image_t raw;
vpx_codec_err_t res;
VpxVideoInfo info = {0};
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
// Test encoder/decoder mismatch.
int test_decode = 1;
// Decoder
vpx_codec_ctx_t dcodec;
unsigned int frame_out = 0;
// The frame number to set reference frame on
unsigned int update_frame_num = 0;
int mismatch_seen = 0;
const int fps = 30;
const int bitrate = 500;
const char *width_arg = NULL;
const char *height_arg = NULL;
const char *infile_arg = NULL;
const char *outfile_arg = NULL;
unsigned int limit = 0;
exec_name = argv[0];
if (argc < 6)
die("Invalid number of arguments");
width_arg = argv[1];
height_arg = argv[2];
infile_arg = argv[3];
outfile_arg = argv[4];
encoder = get_vpx_encoder_by_name("vp9");
if (!encoder)
die("Unsupported codec.");
update_frame_num = atoi(argv[5]);
// In VP9, the reference buffers (cm->buffer_pool->frame_bufs[i].buf) are
// allocated while calling vpx_codec_encode(), thus, setting reference for
// 1st frame isn't supported.
if (update_frame_num <= 1)
die("Couldn't parse frame number '%s'\n", argv[5]);
if (argc > 6) {
limit = atoi(argv[6]);
if (update_frame_num > limit)
die("Update frame number couldn't larger than limit\n");
}
info.codec_fourcc = encoder->fourcc;
info.frame_width = strtol(width_arg, NULL, 0);
info.frame_height = strtol(height_arg, NULL, 0);
info.time_base.numerator = 1;
info.time_base.denominator = fps;
if (info.frame_width <= 0 ||
info.frame_height <= 0 ||
(info.frame_width % 2) != 0 ||
(info.frame_height % 2) != 0) {
die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
}
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
info.frame_height, 1)) {
die("Failed to allocate image.");
}
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res)
die_codec(&ecodec, "Failed to get default codec config.");
cfg.g_w = info.frame_width;
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
cfg.rc_target_bitrate = bitrate;
cfg.g_lag_in_frames = 3;
writer = vpx_video_writer_open(outfile_arg, kContainerIVF, &info);
if (!writer)
die("Failed to open %s for writing.", outfile_arg);
if (!(infile = fopen(infile_arg, "rb")))
die("Failed to open %s for reading.", infile_arg);
if (vpx_codec_enc_init(&ecodec, encoder->codec_interface(), &cfg, 0))
die_codec(&ecodec, "Failed to initialize encoder");
// Disable alt_ref.
if (vpx_codec_control(&ecodec, VP8E_SET_ENABLEAUTOALTREF, 0))
die_codec(&ecodec, "Failed to set enable auto alt ref");
if (test_decode) {
const VpxInterface *decoder = get_vpx_decoder_by_name("vp9");
if (vpx_codec_dec_init(&dcodec, decoder->codec_interface(), NULL, 0))
die_codec(&dcodec, "Failed to initialize decoder.");
}
// Encode frames.
while (vpx_img_read(&raw, infile)) {
if (limit && frame_in >= limit)
break;
if (update_frame_num > 1 && frame_out + 1 == update_frame_num) {
vpx_ref_frame_t ref;
ref.frame_type = VP8_LAST_FRAME;
ref.img = raw;
// Set reference frame in encoder.
if (vpx_codec_control(&ecodec, VP8_SET_REFERENCE, &ref))
die_codec(&ecodec, "Failed to set reference frame");
printf(" <SET_REF>");
// If set_reference in decoder is commented out, the enc/dec mismatch
// would be seen.
if (test_decode) {
if (vpx_codec_control(&dcodec, VP8_SET_REFERENCE, &ref))
die_codec(&dcodec, "Failed to set reference frame");
}
}
encode_frame(&ecodec, &cfg, &raw, frame_in, writer, test_decode,
&dcodec, &frame_out, &mismatch_seen);
frame_in++;
if (mismatch_seen)
break;
}
// Flush encoder.
if (!mismatch_seen)
while (encode_frame(&ecodec, &cfg, NULL, frame_in, writer, test_decode,
&dcodec, &frame_out, &mismatch_seen)) {}
printf("\n");
fclose(infile);
printf("Processed %d frames.\n", frame_out);
if (test_decode) {
if (!mismatch_seen)
printf("Encoder/decoder results are matching.\n");
else
printf("Encoder/decoder results are NOT matching.\n");
}
if (test_decode)
if (vpx_codec_destroy(&dcodec))
die_codec(&dcodec, "Failed to destroy decoder");
vpx_img_free(&raw);
if (vpx_codec_destroy(&ecodec))
die_codec(&ecodec, "Failed to destroy encoder.");
vpx_video_writer_close(writer);
return EXIT_SUCCESS;
}
// TODO(tomfinegan): Improve command line parsing and add args for bitrate/fps.
int main(int argc, char **argv) {
FILE *infile = NULL;
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_count = 0;
vpx_image_t raw;
vpx_codec_err_t res;
VpxVideoInfo info = { 0, 0, 0, { 0, 0 } };
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
const int fps = 30;
const int bitrate = 200;
int keyframe_interval = 0;
int max_frames = 0;
int frames_encoded = 0;
const char *codec_arg = NULL;
const char *width_arg = NULL;
const char *height_arg = NULL;
const char *infile_arg = NULL;
const char *outfile_arg = NULL;
const char *keyframe_interval_arg = NULL;
exec_name = argv[0];
if (argc != 9) die("Invalid number of arguments");
codec_arg = argv[1];
width_arg = argv[2];
height_arg = argv[3];
infile_arg = argv[4];
outfile_arg = argv[5];
keyframe_interval_arg = argv[6];
max_frames = (int)strtol(argv[8], NULL, 0);
encoder = get_vpx_encoder_by_name(codec_arg);
if (!encoder) die("Unsupported codec.");
info.codec_fourcc = encoder->fourcc;
info.frame_width = (int)strtol(width_arg, NULL, 0);
info.frame_height = (int)strtol(height_arg, NULL, 0);
info.time_base.numerator = 1;
info.time_base.denominator = fps;
if (info.frame_width <= 0 || info.frame_height <= 0 ||
(info.frame_width % 2) != 0 || (info.frame_height % 2) != 0) {
die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
}
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
info.frame_height, 1)) {
die("Failed to allocate image.");
}
keyframe_interval = (int)strtol(keyframe_interval_arg, NULL, 0);
if (keyframe_interval < 0) die("Invalid keyframe interval value.");
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) die_codec(&codec, "Failed to get default codec config.");
cfg.g_w = info.frame_width;
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
cfg.rc_target_bitrate = bitrate;
cfg.g_error_resilient = (vpx_codec_er_flags_t)strtoul(argv[7], NULL, 0);
writer = vpx_video_writer_open(outfile_arg, kContainerIVF, &info);
if (!writer) die("Failed to open %s for writing.", outfile_arg);
if (!(infile = fopen(infile_arg, "rb")))
die("Failed to open %s for reading.", infile_arg);
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
die_codec(&codec, "Failed to initialize encoder");
// Encode frames.
while (vpx_img_read(&raw, infile)) {
int flags = 0;
if (keyframe_interval > 0 && frame_count % keyframe_interval == 0)
flags |= VPX_EFLAG_FORCE_KF;
encode_frame(&codec, &raw, frame_count++, flags, writer);
frames_encoded++;
if (max_frames > 0 && frames_encoded >= max_frames) break;
}
// Flush encoder.
while (encode_frame(&codec, NULL, -1, 0, writer)) {
}
printf("\n");
fclose(infile);
printf("Processed %d frames.\n", frame_count);
vpx_img_free(&raw);
if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec.");
vpx_video_writer_close(writer);
return EXIT_SUCCESS;
}
int main(int argc, char **argv) {
VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = {NULL};
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_cnt = 0;
vpx_image_t raw;
vpx_codec_err_t res;
unsigned int width;
unsigned int height;
int speed;
int frame_avail;
int got_data;
int flags = 0;
unsigned int i;
int pts = 0; // PTS starts at 0.
int frame_duration = 1; // 1 timebase tick per frame.
int layering_mode = 0;
int layer_flags[VPX_TS_MAX_PERIODICITY] = {0};
int flag_periodicity = 1;
vpx_svc_layer_id_t layer_id = {0, 0};
const VpxInterface *encoder = NULL;
FILE *infile = NULL;
struct RateControlMetrics rc;
int64_t cx_time = 0;
const int min_args_base = 11;
#if CONFIG_VP9_HIGHBITDEPTH
vpx_bit_depth_t bit_depth = VPX_BITS_8;
int input_bit_depth = 8;
const int min_args = min_args_base + 1;
#else
const int min_args = min_args_base;
#endif // CONFIG_VP9_HIGHBITDEPTH
double sum_bitrate = 0.0;
double sum_bitrate2 = 0.0;
double framerate = 30.0;
exec_name = argv[0];
// Check usage and arguments.
if (argc < min_args) {
#if CONFIG_VP9_HIGHBITDEPTH
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n", argv[0]);
#else
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<Rate_0> ... <Rate_nlayers-1> \n", argv[0]);
#endif // CONFIG_VP9_HIGHBITDEPTH
}
encoder = get_vpx_encoder_by_name(argv[3]);
if (!encoder)
die("Unsupported codec.");
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
width = strtol(argv[4], NULL, 0);
height = strtol(argv[5], NULL, 0);
if (width < 16 || width % 2 || height < 16 || height % 2) {
die("Invalid resolution: %d x %d", width, height);
}
layering_mode = strtol(argv[10], NULL, 0);
if (layering_mode < 0 || layering_mode > 12) {
die("Invalid layering mode (0..12) %s", argv[10]);
}
if (argc != min_args + mode_to_num_layers[layering_mode]) {
die("Invalid number of arguments");
}
#if CONFIG_VP9_HIGHBITDEPTH
switch (strtol(argv[argc-1], NULL, 0)) {
case 8:
bit_depth = VPX_BITS_8;
input_bit_depth = 8;
break;
case 10:
bit_depth = VPX_BITS_10;
input_bit_depth = 10;
break;
case 12:
bit_depth = VPX_BITS_12;
input_bit_depth = 12;
break;
default:
die("Invalid bit depth (8, 10, 12) %s", argv[argc-1]);
}
if (!vpx_img_alloc(&raw,
bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 :
VPX_IMG_FMT_I42016,
width, height, 32)) {
die("Failed to allocate image", width, height);
}
#else
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
die("Failed to allocate image", width, height);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Populate encoder configuration.
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) {
printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
return EXIT_FAILURE;
}
// Update the default configuration with our settings.
cfg.g_w = width;
cfg.g_h = height;
#if CONFIG_VP9_HIGHBITDEPTH
if (bit_depth != VPX_BITS_8) {
cfg.g_bit_depth = bit_depth;
cfg.g_input_bit_depth = input_bit_depth;
cfg.g_profile = 2;
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Timebase format e.g. 30fps: numerator=1, demoninator = 30.
cfg.g_timebase.num = strtol(argv[6], NULL, 0);
cfg.g_timebase.den = strtol(argv[7], NULL, 0);
speed = strtol(argv[8], NULL, 0);
if (speed < 0) {
die("Invalid speed setting: must be positive");
}
for (i = min_args_base;
(int)i < min_args_base + mode_to_num_layers[layering_mode];
++i) {
cfg.ts_target_bitrate[i - 11] = strtol(argv[i], NULL, 0);
}
// Real time parameters.
cfg.rc_dropframe_thresh = strtol(argv[9], NULL, 0);
cfg.rc_end_usage = VPX_CBR;
cfg.rc_resize_allowed = 0;
cfg.rc_min_quantizer = 2;
cfg.rc_max_quantizer = 56;
cfg.rc_undershoot_pct = 50;
cfg.rc_overshoot_pct = 50;
cfg.rc_buf_initial_sz = 500;
cfg.rc_buf_optimal_sz = 600;
cfg.rc_buf_sz = 1000;
// Enable error resilient mode.
cfg.g_error_resilient = 1;
cfg.g_lag_in_frames = 0;
cfg.kf_mode = VPX_KF_AUTO;
// Disable automatic keyframe placement.
cfg.kf_min_dist = cfg.kf_max_dist = 3000;
set_temporal_layer_pattern(layering_mode,
&cfg,
layer_flags,
&flag_periodicity);
set_rate_control_metrics(&rc, &cfg);
// Target bandwidth for the whole stream.
// Set to ts_target_bitrate for highest layer (total bitrate).
cfg.rc_target_bitrate = cfg.ts_target_bitrate[cfg.ts_number_layers - 1];
// Open input file.
if (!(infile = fopen(argv[1], "rb"))) {
die("Failed to open %s for reading", argv[1]);
}
framerate = cfg.g_timebase.den / cfg.g_timebase.num;
// Open an output file for each stream.
for (i = 0; i < cfg.ts_number_layers; ++i) {
char file_name[PATH_MAX];
VpxVideoInfo info;
info.codec_fourcc = encoder->fourcc;
info.frame_width = cfg.g_w;
info.frame_height = cfg.g_h;
info.time_base.numerator = cfg.g_timebase.num;
info.time_base.denominator = cfg.g_timebase.den;
snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
if (!outfile[i])
die("Failed to open %s for writing", file_name);
assert(outfile[i] != NULL);
}
// No spatial layers in this encoder.
cfg.ss_number_layers = 1;
// Initialize codec.
#if CONFIG_VP9_HIGHBITDEPTH
if (vpx_codec_enc_init(
&codec, encoder->codec_interface(), &cfg,
bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
#else
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
#endif // CONFIG_VP9_HIGHBITDEPTH
die_codec(&codec, "Failed to initialize encoder");
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kDenoiserOnYOnly);
} else if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0);
vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, 0);
if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1: 0)) {
die_codec(&codec, "Failed to set SVC");
}
}
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP8E_SET_TOKEN_PARTITIONS, 1);
// This controls the maximum target size of the key frame.
// For generating smaller key frames, use a smaller max_intra_size_pct
// value, like 100 or 200.
{
const int max_intra_size_pct = 900;
vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT,
max_intra_size_pct);
}
frame_avail = 1;
while (frame_avail || got_data) {
struct vpx_usec_timer timer;
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t *pkt;
// Update the temporal layer_id. No spatial layers in this test.
layer_id.spatial_layer_id = 0;
layer_id.temporal_layer_id =
cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
}
flags = layer_flags[frame_cnt % flag_periodicity];
frame_avail = vpx_img_read(&raw, infile);
if (frame_avail)
++rc.layer_input_frames[layer_id.temporal_layer_id];
vpx_usec_timer_start(&timer);
if (vpx_codec_encode(&codec, frame_avail? &raw : NULL, pts, 1, flags,
VPX_DL_REALTIME)) {
die_codec(&codec, "Failed to encode frame");
}
vpx_usec_timer_mark(&timer);
cx_time += vpx_usec_timer_elapsed(&timer);
// Reset KF flag.
if (layering_mode != 7) {
layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
}
got_data = 0;
while ( (pkt = vpx_codec_get_cx_data(&codec, &iter)) ) {
got_data = 1;
switch (pkt->kind) {
case VPX_CODEC_CX_FRAME_PKT:
for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
i < cfg.ts_number_layers; ++i) {
vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf,
pkt->data.frame.sz, pts);
++rc.layer_tot_enc_frames[i];
rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz;
// Keep count of rate control stats per layer (for non-key frames).
if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] &&
!(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz;
rc.layer_avg_rate_mismatch[i] +=
fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) /
rc.layer_pfb[i];
++rc.layer_enc_frames[i];
}
}
// Update for short-time encoding bitrate states, for moving window
// of size rc->window, shifted by rc->window / 2.
// Ignore first window segment, due to key frame.
if (frame_cnt > rc.window_size) {
sum_bitrate += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
if (frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
rc.variance_st_encoding_bitrate +=
(sum_bitrate / rc.window_size) *
(sum_bitrate / rc.window_size);
sum_bitrate = 0.0;
}
}
// Second shifted window.
if (frame_cnt > rc.window_size + rc.window_size / 2) {
sum_bitrate2 += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
if (frame_cnt > 2 * rc.window_size &&
frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
rc.variance_st_encoding_bitrate +=
(sum_bitrate2 / rc.window_size) *
(sum_bitrate2 / rc.window_size);
sum_bitrate2 = 0.0;
}
}
break;
default:
break;
}
}
++frame_cnt;
pts += frame_duration;
}
fclose(infile);
printout_rate_control_summary(&rc, &cfg, frame_cnt);
printf("\n");
printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
frame_cnt,
1000 * (float)cx_time / (double)(frame_cnt * 1000000),
1000000 * (double)frame_cnt / (double)cx_time);
if (vpx_codec_destroy(&codec))
die_codec(&codec, "Failed to destroy codec");
// Try to rewrite the output file headers with the actual frame count.
for (i = 0; i < cfg.ts_number_layers; ++i)
vpx_video_writer_close(outfile[i]);
vpx_img_free(&raw);
return EXIT_SUCCESS;
}
int main(int argc, char **argv) {
FILE *infile = NULL;
VpxVideoWriter *writer = NULL;
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
vpx_image_t raw;
vpx_codec_err_t res;
vpx_fixed_buf_t stats = {0};
VpxVideoInfo info = {0};
const VpxInterface *encoder = NULL;
int pass;
const int fps = 30; // TODO(dkovalev) add command line argument
const int bitrate = 200; // kbit/s TODO(dkovalev) add command line argument
const char *const codec_arg = argv[1];
const char *const width_arg = argv[2];
const char *const height_arg = argv[3];
const char *const infile_arg = argv[4];
const char *const outfile_arg = argv[5];
exec_name = argv[0];
if (argc != 6)
die("Invalid number of arguments.");
encoder = get_vpx_encoder_by_name(codec_arg);
if (!encoder)
die("Unsupported codec.");
info.codec_fourcc = encoder->fourcc;
info.time_base.numerator = 1;
info.time_base.denominator = fps;
info.frame_width = strtol(width_arg, NULL, 0);
info.frame_height = strtol(height_arg, NULL, 0);
if (info.frame_width <= 0 ||
info.frame_height <= 0 ||
(info.frame_width % 2) != 0 ||
(info.frame_height % 2) != 0) {
die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
}
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
info.frame_height, 1)) {
die("Failed to allocate image", info.frame_width, info.frame_height);
}
writer = vpx_video_writer_open(outfile_arg, kContainerIVF, &info);
if (!writer)
die("Failed to open %s for writing", outfile_arg);
printf("Using %s\n", vpx_codec_iface_name(encoder->interface()));
res = vpx_codec_enc_config_default(encoder->interface(), &cfg, 0);
if (res)
die_codec(&codec, "Failed to get default codec config.");
cfg.g_w = info.frame_width;
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
cfg.rc_target_bitrate = bitrate;
for (pass = 0; pass < 2; ++pass) {
int frame_count = 0;
if (pass == 0) {
cfg.g_pass = VPX_RC_FIRST_PASS;
} else {
cfg.g_pass = VPX_RC_LAST_PASS;
cfg.rc_twopass_stats_in = stats;
}
if (!(infile = fopen(infile_arg, "rb")))
die("Failed to open %s for reading", infile_arg);
if (vpx_codec_enc_init(&codec, encoder->interface(), &cfg, 0))
die_codec(&codec, "Failed to initialize encoder");
while (vpx_img_read(&raw, infile)) {
++frame_count;
if (pass == 0) {
get_frame_stats(&codec, &raw, frame_count, 1, 0, VPX_DL_BEST_QUALITY,
&stats);
} else {
encode_frame(&codec, &raw, frame_count, 1, 0, VPX_DL_BEST_QUALITY,
writer);
}
}
if (pass == 0) {
get_frame_stats(&codec, NULL, frame_count, 1, 0, VPX_DL_BEST_QUALITY,
&stats);
} else {
printf("\n");
}
fclose(infile);
printf("Pass %d complete. Processed %d frames.\n", pass + 1, frame_count);
if (vpx_codec_destroy(&codec))
die_codec(&codec, "Failed to destroy codec.");
}
vpx_img_free(&raw);
free(stats.buf);
vpx_video_writer_close(writer);
return EXIT_SUCCESS;
}
int main(int argc, char **argv) {
FILE *infile = NULL;
vpx_codec_ctx_t codec = {0};
vpx_codec_enc_cfg_t cfg = {0};
int frame_count = 0;
vpx_image_t raw = {0};
vpx_codec_err_t res;
VpxVideoInfo info = {0};
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
const int fps = 2; // TODO(dkovalev) add command line argument
const int bitrate = 200; // kbit/s TODO(dkovalev) add command line argument
exec_name = argv[0];
if (argc != 5)
die("Invalid number of arguments");
encoder = get_vpx_encoder_by_name("vp8"); // only vp8 for now
if (!encoder)
die("Unsupported codec.");
info.codec_fourcc = encoder->fourcc;
info.frame_width = strtol(argv[1], NULL, 0);
info.frame_height = strtol(argv[2], NULL, 0);
info.time_base.numerator = 1;
info.time_base.denominator = fps;
if (info.frame_width <= 0 ||
info.frame_height <= 0 ||
(info.frame_width % 2) != 0 ||
(info.frame_height % 2) != 0) {
die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
}
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
info.frame_height, 1)) {
die("Failed to allocate image.");
}
printf("Using %s\n", vpx_codec_iface_name(encoder->interface()));
res = vpx_codec_enc_config_default(encoder->interface(), &cfg, 0);
if (res)
die_codec(&codec, "Failed to get default codec config.");
cfg.g_w = info.frame_width;
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
cfg.rc_target_bitrate = bitrate;
writer = vpx_video_writer_open(argv[4], kContainerIVF, &info);
if (!writer)
die("Failed to open %s for writing.", argv[4]);
if (!(infile = fopen(argv[3], "rb")))
die("Failed to open %s for reading.", argv[3]);
if (vpx_codec_enc_init(&codec, encoder->interface(), &cfg, 0))
die_codec(&codec, "Failed to initialize encoder");
while (vpx_img_read(&raw, infile)) {
++frame_count;
if (frame_count == 22) {
set_roi_map(&cfg, &codec);
} else if (frame_count == 33) {
set_active_map(&cfg, &codec);
} else if (frame_count == 44) {
unset_active_map(&cfg, &codec);
}
encode_frame(&codec, &raw, frame_count, writer);
}
encode_frame(&codec, NULL, -1, writer);
printf("\n");
fclose(infile);
printf("Processed %d frames.\n", frame_count);
vpx_img_free(&raw);
if (vpx_codec_destroy(&codec))
die_codec(&codec, "Failed to destroy codec.");
vpx_video_writer_close(writer);
return EXIT_SUCCESS;
}
int main(int argc, char **argv) {
FILE *infile = NULL;
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_count = 0;
vpx_image_t raw;
vpx_codec_err_t res;
VpxVideoInfo info;
VpxVideoWriter *writer = NULL;
const VpxInterface *encoder = NULL;
int update_frame_num = 0;
const int fps = 30; // TODO(dkovalev) add command line argument
const int bitrate = 200; // kbit/s TODO(dkovalev) add command line argument
vp8_zero(codec);
vp8_zero(cfg);
vp8_zero(info);
exec_name = argv[0];
if (argc != 6) die("Invalid number of arguments");
// TODO(dkovalev): add vp9 support and rename the file accordingly
encoder = get_vpx_encoder_by_name("vp8");
if (!encoder) die("Unsupported codec.");
update_frame_num = atoi(argv[5]);
if (!update_frame_num) die("Couldn't parse frame number '%s'\n", argv[5]);
info.codec_fourcc = encoder->fourcc;
info.frame_width = (int)strtol(argv[1], NULL, 0);
info.frame_height = (int)strtol(argv[2], NULL, 0);
info.time_base.numerator = 1;
info.time_base.denominator = fps;
if (info.frame_width <= 0 || info.frame_height <= 0 ||
(info.frame_width % 2) != 0 || (info.frame_height % 2) != 0) {
die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
}
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
info.frame_height, 1)) {
die("Failed to allocate image.");
}
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) die_codec(&codec, "Failed to get default codec config.");
cfg.g_w = info.frame_width;
cfg.g_h = info.frame_height;
cfg.g_timebase.num = info.time_base.numerator;
cfg.g_timebase.den = info.time_base.denominator;
cfg.rc_target_bitrate = bitrate;
writer = vpx_video_writer_open(argv[4], kContainerIVF, &info);
if (!writer) die("Failed to open %s for writing.", argv[4]);
if (!(infile = fopen(argv[3], "rb")))
die("Failed to open %s for reading.", argv[3]);
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
die_codec(&codec, "Failed to initialize encoder");
// Encode frames.
while (vpx_img_read(&raw, infile)) {
if (frame_count + 1 == update_frame_num) {
vpx_ref_frame_t ref;
ref.frame_type = VP8_LAST_FRAME;
ref.img = raw;
if (vpx_codec_control(&codec, VP8_SET_REFERENCE, &ref))
die_codec(&codec, "Failed to set reference frame");
}
encode_frame(&codec, &raw, frame_count++, writer);
}
// Flush encoder.
while (encode_frame(&codec, NULL, -1, writer)) {
}
printf("\n");
fclose(infile);
printf("Processed %d frames.\n", frame_count);
vpx_img_free(&raw);
if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec.");
vpx_video_writer_close(writer);
return EXIT_SUCCESS;
}