i am having problems understanding how the audio part of the sdl library works
now, i know that when you initialize it, you have to specify the frequency and a >>callback<< function, which i think is then called automatically at the given frequency.
can anyone who worked with the sdl library write a simple example that would use sdl_audio to generate a 440 hz square wave (since it is the simplest waveform) at a sampling frequency of 44000 hz?
The Introduction to SDL (2011 cached version: 2) has got a neat example of using SDL Sound library that should get you started: http://www.libsdl.org/intro.en/usingsound.html
EDIT: Here is a working program that does what you asked for. I modified a bit the code found here: http://www.dgames.org/beep-sound-with-sdl/
#include <SDL/SDL.h>
#include <SDL/SDL_audio.h>
#include <queue>
#include <cmath>
const int AMPLITUDE = 28000;
const int FREQUENCY = 44100;
struct BeepObject
{
double freq;
int samplesLeft;
};
class Beeper
{
private:
double v;
std::queue<BeepObject> beeps;
public:
Beeper();
~Beeper();
void beep(double freq, int duration);
void generateSamples(Sint16 *stream, int length);
void wait();
};
void audio_callback(void*, Uint8*, int);
Beeper::Beeper()
{
SDL_AudioSpec desiredSpec;
desiredSpec.freq = FREQUENCY;
desiredSpec.format = AUDIO_S16SYS;
desiredSpec.channels = 1;
desiredSpec.samples = 2048;
desiredSpec.callback = audio_callback;
desiredSpec.userdata = this;
SDL_AudioSpec obtainedSpec;
// you might want to look for errors here
SDL_OpenAudio(&desiredSpec, &obtainedSpec);
// start play audio
SDL_PauseAudio(0);
}
Beeper::~Beeper()
{
SDL_CloseAudio();
}
void Beeper::generateSamples(Sint16 *stream, int length)
{
int i = 0;
while (i < length) {
if (beeps.empty()) {
while (i < length) {
stream[i] = 0;
i++;
}
return;
}
BeepObject& bo = beeps.front();
int samplesToDo = std::min(i + bo.samplesLeft, length);
bo.samplesLeft -= samplesToDo - i;
while (i < samplesToDo) {
stream[i] = AMPLITUDE * std::sin(v * 2 * M_PI / FREQUENCY);
i++;
v += bo.freq;
}
if (bo.samplesLeft == 0) {
beeps.pop();
}
}
}
void Beeper::beep(double freq, int duration)
{
BeepObject bo;
bo.freq = freq;
bo.samplesLeft = duration * FREQUENCY / 1000;
SDL_LockAudio();
beeps.push(bo);
SDL_UnlockAudio();
}
void Beeper::wait()
{
int size;
do {
SDL_Delay(20);
SDL_LockAudio();
size = beeps.size();
SDL_UnlockAudio();
} while (size > 0);
}
void audio_callback(void *_beeper, Uint8 *_stream, int _length)
{
Sint16 *stream = (Sint16*) _stream;
int length = _length / 2;
Beeper* beeper = (Beeper*) _beeper;
beeper->generateSamples(stream, length);
}
int main(int argc, char* argv[])
{
SDL_Init(SDL_INIT_AUDIO);
int duration = 1000;
double Hz = 440;
Beeper b;
b.beep(Hz, duration);
b.wait();
return 0;
}
Good luck.
A boiled-down variant of the beeper-example, reduced to the bare minimum (with error-handling).
#include <math.h>
#include <SDL.h>
#include <SDL_audio.h>
const int AMPLITUDE = 28000;
const int SAMPLE_RATE = 44100;
void audio_callback(void *user_data, Uint8 *raw_buffer, int bytes)
{
Sint16 *buffer = (Sint16*)raw_buffer;
int length = bytes / 2; // 2 bytes per sample for AUDIO_S16SYS
int &sample_nr(*(int*)user_data);
for(int i = 0; i < length; i++, sample_nr++)
{
double time = (double)sample_nr / (double)SAMPLE_RATE;
buffer[i] = (Sint16)(AMPLITUDE * sin(2.0f * M_PI * 441.0f * time)); // render 441 HZ sine wave
}
}
int main(int argc, char *argv[])
{
if(SDL_Init(SDL_INIT_AUDIO) != 0) SDL_Log("Failed to initialize SDL: %s", SDL_GetError());
int sample_nr = 0;
SDL_AudioSpec want;
want.freq = SAMPLE_RATE; // number of samples per second
want.format = AUDIO_S16SYS; // sample type (here: signed short i.e. 16 bit)
want.channels = 1; // only one channel
want.samples = 2048; // buffer-size
want.callback = audio_callback; // function SDL calls periodically to refill the buffer
want.userdata = &sample_nr; // counter, keeping track of current sample number
SDL_AudioSpec have;
if(SDL_OpenAudio(&want, &have) != 0) SDL_LogError(SDL_LOG_CATEGORY_AUDIO, "Failed to open audio: %s", SDL_GetError());
if(want.format != have.format) SDL_LogError(SDL_LOG_CATEGORY_AUDIO, "Failed to get the desired AudioSpec");
SDL_PauseAudio(0); // start playing sound
SDL_Delay(1000); // wait while sound is playing
SDL_PauseAudio(1); // stop playing sound
SDL_CloseAudio();
return 0;
}
SDL 2 C example
The following code produces a sinusoidal sound, it is adapted from: https://codereview.stackexchange.com/questions/41086/play-some-sine-waves-with-sdl2
main.c
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <SDL2/SDL.h>
const double ChromaticRatio = 1.059463094359295264562;
const double Tao = 6.283185307179586476925;
Uint32 sampleRate = 48000;
Uint32 frameRate = 60;
Uint32 floatStreamLength = 1024;
Uint32 samplesPerFrame;
Uint32 msPerFrame;
double practicallySilent = 0.001;
Uint32 audioBufferLength = 48000;
float *audioBuffer;
SDL_atomic_t audioCallbackLeftOff;
Sint32 audioMainLeftOff;
Uint8 audioMainAccumulator;
SDL_AudioDeviceID AudioDevice;
SDL_AudioSpec audioSpec;
SDL_Event event;
SDL_bool running = SDL_TRUE;
typedef struct {
float *waveform;
Uint32 waveformLength;
double volume;
double pan;
double frequency;
double phase;
} voice;
void speak(voice *v) {
float sample;
Uint32 sourceIndex;
double phaseIncrement = v->frequency/sampleRate;
Uint32 i;
if (v->volume > practicallySilent) {
for (i = 0; (i + 1) < samplesPerFrame; i += 2) {
v->phase += phaseIncrement;
if (v->phase > 1)
v->phase -= 1;
sourceIndex = v->phase*v->waveformLength;
sample = v->waveform[sourceIndex]*v->volume;
audioBuffer[audioMainLeftOff+i] += sample*(1-v->pan);
audioBuffer[audioMainLeftOff+i+1] += sample*v->pan;
}
}
else {
for (i=0; i<samplesPerFrame; i+=1)
audioBuffer[audioMainLeftOff+i] = 0;
}
audioMainAccumulator++;
}
double getFrequency(double pitch) {
return pow(ChromaticRatio, pitch-57)*440;
}
int getWaveformLength(double pitch) {
return sampleRate / getFrequency(pitch)+0.5f;
}
void buildSineWave(float *data, Uint32 length) {
Uint32 i;
for (i=0; i < length; i++)
data[i] = sin(i*(Tao/length));
}
void logSpec(SDL_AudioSpec *as) {
printf(
" freq______%5d\n"
" format____%5d\n"
" channels__%5d\n"
" silence___%5d\n"
" samples___%5d\n"
" size______%5d\n\n",
(int) as->freq,
(int) as->format,
(int) as->channels,
(int) as->silence,
(int) as->samples,
(int) as->size
);
}
void logVoice(voice *v) {
printf(
" waveformLength__%d\n"
" volume__________%f\n"
" pan_____________%f\n"
" frequency_______%f\n"
" phase___________%f\n",
v->waveformLength,
v->volume,
v->pan,
v->frequency,
v->phase
);
}
void logWavedata(float *floatStream, Uint32 floatStreamLength, Uint32 increment) {
printf("\n\nwaveform data:\n\n");
Uint32 i=0;
for (i = 0; i < floatStreamLength; i += increment)
printf("%4d:%2.16f\n", i, floatStream[i]);
printf("\n\n");
}
void audioCallback(void *unused, Uint8 *byteStream, int byteStreamLength) {
float* floatStream = (float*) byteStream;
Sint32 localAudioCallbackLeftOff = SDL_AtomicGet(&audioCallbackLeftOff);
Uint32 i;
for (i = 0; i < floatStreamLength; i++) {
floatStream[i] = audioBuffer[localAudioCallbackLeftOff];
localAudioCallbackLeftOff++;
if (localAudioCallbackLeftOff == audioBufferLength)
localAudioCallbackLeftOff = 0;
}
SDL_AtomicSet(&audioCallbackLeftOff, localAudioCallbackLeftOff);
}
int init(void) {
SDL_Init(SDL_INIT_AUDIO | SDL_INIT_TIMER);
SDL_AudioSpec want;
SDL_zero(want);
want.freq = sampleRate;
want.format = AUDIO_F32;
want.channels = 2;
want.samples = floatStreamLength;
want.callback = audioCallback;
AudioDevice = SDL_OpenAudioDevice(NULL, 0, &want, &audioSpec, SDL_AUDIO_ALLOW_FORMAT_CHANGE);
if (AudioDevice == 0) {
printf("\nFailed to open audio: %s\n", SDL_GetError());
return 1;
}
printf("want:\n");
logSpec(&want);
printf("audioSpec:\n");
logSpec(&audioSpec);
if (audioSpec.format != want.format) {
printf("\nCouldn't get Float32 audio format.\n");
return 2;
}
sampleRate = audioSpec.freq;
floatStreamLength = audioSpec.size / 4;
samplesPerFrame = sampleRate / frameRate;
msPerFrame = 1000 / frameRate;
audioMainLeftOff = samplesPerFrame * 8;
SDL_AtomicSet(&audioCallbackLeftOff, 0);
if (audioBufferLength % samplesPerFrame)
audioBufferLength += samplesPerFrame - (audioBufferLength % samplesPerFrame);
audioBuffer = malloc(sizeof(float) * audioBufferLength);
return 0;
}
int onExit(void) {
SDL_CloseAudioDevice(AudioDevice);
SDL_Quit();
return 0;
}
int main(int argc, char *argv[]) {
float syncCompensationFactor = 0.0016;
Sint32 mainAudioLead;
Uint32 i;
voice testVoiceA;
voice testVoiceB;
voice testVoiceC;
testVoiceA.volume = 1;
testVoiceB.volume = 1;
testVoiceC.volume = 1;
testVoiceA.pan = 0.5;
testVoiceB.pan = 0;
testVoiceC.pan = 1;
testVoiceA.phase = 0;
testVoiceB.phase = 0;
testVoiceC.phase = 0;
testVoiceA.frequency = getFrequency(45);
testVoiceB.frequency = getFrequency(49);
testVoiceC.frequency = getFrequency(52);
Uint16 C0waveformLength = getWaveformLength(0);
testVoiceA.waveformLength = C0waveformLength;
testVoiceB.waveformLength = C0waveformLength;
testVoiceC.waveformLength = C0waveformLength;
float sineWave[C0waveformLength];
buildSineWave(sineWave, C0waveformLength);
testVoiceA.waveform = sineWave;
testVoiceB.waveform = sineWave;
testVoiceC.waveform = sineWave;
if (init())
return 1;
SDL_Delay(42);
SDL_PauseAudioDevice(AudioDevice, 0);
while (running) {
while (SDL_PollEvent(&event) != 0) {
if (event.type == SDL_QUIT) {
running = SDL_FALSE;
}
}
for (i = 0; i < samplesPerFrame; i++)
audioBuffer[audioMainLeftOff+i] = 0;
speak(&testVoiceA);
speak(&testVoiceB);
speak(&testVoiceC);
if (audioMainAccumulator > 1) {
for (i=0; i<samplesPerFrame; i++) {
audioBuffer[audioMainLeftOff+i] /= audioMainAccumulator;
}
}
audioMainAccumulator = 0;
audioMainLeftOff += samplesPerFrame;
if (audioMainLeftOff == audioBufferLength)
audioMainLeftOff = 0;
mainAudioLead = audioMainLeftOff - SDL_AtomicGet(&audioCallbackLeftOff);
if (mainAudioLead < 0)
mainAudioLead += audioBufferLength;
if (mainAudioLead < floatStreamLength)
printf("An audio collision may have occured!\n");
SDL_Delay(mainAudioLead * syncCompensationFactor);
}
onExit();
return 0;
}
Compile and run:
gcc -ggdb3 -O3 -std=c99 -Wall -Wextra -pedantic -o main.out main.c -lSDL2 -lm
./main.out
Should be easy to turn this into a simple piano with: https://github.com/cirosantilli/cpp-cheat/blob/f734a2e76fbcfc67f707ae06be7a2a2ef5db47d1/c/interactive/audio_gen.c#L44
For wav manipulation, also check the official examples:
http://hg.libsdl.org/SDL/file/e12c38730512/test/testresample.c
http://hg.libsdl.org/SDL/file/e12c38730512/test/loopwave.c
Tested on Ubuntu 19.10, SDL 2.0.10.
This is a minimal example of how to play a sine wave in SDL2.
Make sure to call SDL_Init(SDL_INIT_AUDIO) before creating an instance of Sound.
Sound.h
#include <cstdint>
#include <SDL2/SDL.h>
class Sound
{
public:
Sound();
~Sound();
void play();
void stop();
const double m_sineFreq;
const double m_sampleFreq;
const double m_samplesPerSine;
uint32_t m_samplePos;
private:
static void SDLAudioCallback(void *data, Uint8 *buffer, int length);
SDL_AudioDeviceID m_device;
};
Sound.cpp
#include "Sound.h"
#include <cmath>
#include <iostream>
Sound::Sound()
: m_sineFreq(1000),
m_sampleFreq(44100),
m_samplesPerSine(m_sampleFreq / m_sineFreq),
m_samplePos(0)
{
SDL_AudioSpec wantSpec, haveSpec;
SDL_zero(wantSpec);
wantSpec.freq = m_sampleFreq;
wantSpec.format = AUDIO_U8;
wantSpec.channels = 1;
wantSpec.samples = 2048;
wantSpec.callback = SDLAudioCallback;
wantSpec.userdata = this;
m_device = SDL_OpenAudioDevice(NULL, 0, &wantSpec, &haveSpec, SDL_AUDIO_ALLOW_FORMAT_CHANGE);
if (m_device == 0)
{
std::cout << "Failed to open audio: " << SDL_GetError() << std::endl;
}
}
Sound::~Sound()
{
SDL_CloseAudioDevice(m_device);
}
void Sound::play()
{
SDL_PauseAudioDevice(m_device, 0);
}
void Sound::stop()
{
SDL_PauseAudioDevice(m_device, 1);
}
void Sound::SDLAudioCallback(void *data, Uint8 *buffer, int length)
{
Sound *sound = reinterpret_cast<Sound*>(data);
for(int i = 0; i < length; ++i)
{
buffer[i] = (std::sin(sound->m_samplePos / sound->m_samplesPerSine * M_PI * 2) + 1) * 127.5;
++sound->m_samplePos;
}
}
Related
I am trying to generate a simple, constant sine tone using SDL_audio. I have a small helper class that can be called to turn the tone on/off, change the frequency, and change the wave shape. I have followed some examples I could find on the web and got the following:
beeper.h
#pragma once
#include <SDL.h>
#include <SDL_audio.h>
#include <cmath>
#include "logger.h"
class Beeper {
private:
//Should there be sound right now
bool soundOn = true;
//Type of wave that should be generated
int waveType = 0;
//Tone that the wave will produce (may or may not be applicable based on wave type)
float waveTone = 440;
//Running index for sampling
float samplingIndex = 0;
//These are useful variables that cannot be changed outside of this file:
//Volume
const Sint16 amplitude = 32000;
//Sampling rate
const int samplingRate = 44100;
//Buffer size
const int bufferSize = 1024;
//Samples a sine wave at a given index
float sampleSine(float index);
//Samples a square wave at a given index
float sampleSquare(float index);
public:
//Initializes SDL audio, audio device, and audio specs
void initializeAudio();
//Function called by SDL audio_callback that fills stream with samples
void generateSamples(short* stream, int length);
//Turn sound on or off
void setSoundOn(bool soundOnOrOff);
//Set timbre of tone produced by beeper
void setWaveType(int waveTypeID);
//Set tone (in Hz) produced by beeper
void setWaveTone(int waveHz);
};
beeper.cpp
#include <beeper.h>
void fillBuffer(void* userdata, Uint8* _stream, int len) {
short * stream = reinterpret_cast<short*>(_stream);
int length = len;
Beeper* beeper = (Beeper*)userdata;
beeper->generateSamples(stream, length);
}
void Beeper::initializeAudio() {
SDL_AudioSpec desired, returned;
SDL_AudioDeviceID devID;
SDL_zero(desired);
desired.freq = samplingRate;
desired.format = AUDIO_S16SYS; //16-bit audio
desired.channels = 1;
desired.samples = bufferSize;
desired.callback = &fillBuffer;
desired.userdata = this;
devID = SDL_OpenAudioDevice(SDL_GetAudioDeviceName(0,0), 0, &desired, &returned, SDL_AUDIO_ALLOW_FORMAT_CHANGE);
SDL_PauseAudioDevice(devID, 0);
}
void Beeper::generateSamples(short *stream, int length) {
int samplesToWrite = length / sizeof(short);
for (int i = 0; i < samplesToWrite; i++) {
if (soundOn) {
if (waveType == 0) {
stream[i] = (short)(amplitude * sampleSine(samplingIndex));
}
else if (waveType == 1) {
stream[i] = (short)(amplitude * 0.8 * sampleSquare(samplingIndex));
}
}
else {
stream[i] = 0;
}
//INFO << "Sampling index: " << samplingIndex;
samplingIndex += (waveTone * M_PI * 2) / samplingRate;
//INFO << "Stream input: " << stream[i];
if (samplingIndex >= (M_PI*2)) {
samplingIndex -= M_PI * 2;
}
}
}
void Beeper::setSoundOn(bool soundOnOrOff) {
soundOn = soundOnOrOff;
//if (soundOnOrOff) {
// samplingIndex = 0;
//}
}
void Beeper::setWaveType(int waveTypeID) {
waveType = waveTypeID;
//samplingIndex = 0;
}
void Beeper::setWaveTone(int waveHz) {
waveTone = waveHz;
//samplingIndex = 0;
}
float Beeper::sampleSine(float index) {
double result = sin((index));
//INFO << "Sine result: " << result;
return result;
}
float Beeper::sampleSquare(float index)
{
int unSquaredSin = sin((index));
if (unSquaredSin >= 0) {
return 1;
}
else {
return -1;
}
}
The callback function is being called and the generateSamples function is loading data into the stream, but I cannot hear anything but a very slight click at irregular periods. I have had a look at the data inside the stream and it follows a pattern that I would expect for a scaled sine wave with a 440 Hz frequency. Is there something obvious that I am missing? I did notice that the size of the stream is double what I put when declaring the SDL_AudioSpec and calling SDL_OpenAudioDevice. Why is that?
Answered my own question! When opening the audio device I used the flag SDL_AUDIO_ALLOW_FORMAT_CHANGE which meant that SDL was actually using a float buffer instead of the short buffer that I expected. This was causing issues in a couple of places that were hard to detect (the stream being double the amount of bytes I was expecting should have tipped me off). I changed that parameter in SDL_OpenAudioDevice() to 0 and it worked as expected!
I am trying to play a sin wave sound with SDL2 by using the audio queue on C++. In order to do that, I have created a class "Speaker", which has a pushBeep function that is called every time a beep needs to be generated. I have created an AudioDevice successfully, and it is also successful when I do the QueueAudio to the device (I have checked on the debugger) but I can't seem to get any sound out of it.
I have tried changing the way I generate the samples in numerous ways, also, as I said previously, I have checked that the device is properly opened and the QueueAudio returns 0 for success.
This is the class
Speaker::Speaker()
{
SDL_AudioSpec ds;
ds.freq = Speaker::SPEAKER_FREQUENCY;
ds.format = AUDIO_F32;
ds.channels = 1;
ds.samples = 4096;
ds.callback = NULL;
ds.userdata = this;
SDL_AudioSpec os;
this->dev = SDL_OpenAudioDevice(NULL, 0, &ds, &os, NULL);
std::cout << "DEVICE: " << this->dev << std::endl;
SDL_PauseAudioDevice(this->dev, 0);
}
Speaker::~Speaker()
{
SDL_CloseAudioDevice(this->dev);
}
void Speaker::pushBeep(double freq, int duration) {
int nSamples = duration * Speaker::SPEAKER_FREQUENCY / 1000;
float* samples = new float[nSamples];
double v = 0.0;
for (int idx = 0; idx < nSamples; idx++) {
//float value = (float)Speaker::SPEAKER_AMPLITUDE * std::sin(v * 2 * M_PI / Speaker::SPEAKER_FREQUENCY);
float value = 440.0;
samples[idx] = value;
v += freq;
}
int a = SDL_QueueAudio(this->dev, (void*)samples, nSamples * sizeof(float));
std::cout << a << std::endl;
delete[] samples;
samples = NULL;
}
And this is how I call it
Speaker s;
s.pushBeep(440.0, 1000);
When I try with the sin wave generation code (commented) it gives me a "double to float loss of precision" error. When I use the fixed value (not commented) it does not give the error, but it still does not work.
I expect the program to output the sound.
Couple of things you are missing, or maybe you didn't add to your code snippet. You didn't specify an audio callback so when you call SDL_QueueAudio(); it didn't know what to do with the data I'm pretty sure. And you weren't calling SDL_PauseAudioDevice() in your example with the delay.
#include <math.h>
#include <SDL2/SDL.h>
#include <SDL2/SDL_audio.h>
#include <iostream>
namespace AudioGen
{
const int AMPLITUDE = 1;
const int SAMPLE_RATE = 44000;
// Globals
float *in_buffer;
SDL_atomic_t callback_sample_pos;
SDL_Event event;
SDL_bool running = SDL_TRUE;
/**
* Structure for holding audio metadata such as frequency
*/
struct AudioData
{
int sampleNum;
float frequency;
};
void audio_callback(void *user_data, Uint8 *raw_buffer, int bytes)
{
float *buffer = (float*)raw_buffer;
AudioData &audio_data(*static_cast<AudioData*>(user_data));
int nSamples = bytes / 4; // For F32
std::cout << nSamples << std::endl;
for(int i = 0; i < nSamples; i++, audio_data.sampleNum++)
{
double time = (double)audio_data.sampleNum / (double)SAMPLE_RATE;
buffer[i] = (float)(AMPLITUDE * sin(2.0f * M_PI * audio_data.frequency * time));
}
}
int buffer_length;
void callback(void *user_data, Uint8 *raw_buffer, int bytes)
{
float *buffer = (float*)raw_buffer;
int nSamples = bytes/4;
auto local_sample_pos = SDL_AtomicGet(&callback_sample_pos);
for(int i = 0; i < nSamples; ++i)
{
// Stop running audio if all samples are finished playing
if(buffer_length == local_sample_pos)
{
running = SDL_FALSE;
break;
}
buffer[i] = in_buffer[local_sample_pos];
++local_sample_pos;
}
SDL_AtomicSet(&callback_sample_pos, local_sample_pos);
}
class Speaker
{
public:
Speaker()
{
SDL_Init(SDL_INIT_AUDIO);
SDL_AudioSpec ds;
ds.freq = SAMPLE_RATE;
ds.format = AUDIO_F32;
ds.channels = 1;
ds.samples = 4096;
ds.callback = callback;
ds.userdata = &ad; // metadata for frequency
SDL_AudioSpec os;
dev = SDL_OpenAudioDevice(NULL, 0, &ds, &os, SDL_AUDIO_ALLOW_FORMAT_CHANGE);
}
~Speaker()
{
SDL_CloseAudioDevice(dev);
SDL_Quit();
}
void pushBeep(float frequency, int duration)
{
ad.frequency = frequency; // set the frequency for the beep
SDL_PauseAudioDevice(dev, 0);
SDL_Delay(duration); // wait while sound is playing
SDL_PauseAudioDevice(dev, 1);
}
void pushBeep2(float frequency, int duration )
{
int nSamples = duration * SAMPLE_RATE / 1000;
in_buffer = new float[nSamples];
buffer_length = nSamples;
for (int idx = 0; idx < nSamples; idx++) {
double time = (double)idx / (double)SAMPLE_RATE;
in_buffer[idx] = (float)(AMPLITUDE * std::sin(2.0f * M_PI * frequency * time));
}
SDL_QueueAudio(dev, in_buffer, nSamples * sizeof(float));
SDL_PauseAudioDevice(dev, 0);
while(running){
while(SDL_PollEvent(&event)!=0);
}
delete[] in_buffer;
}
private:
SDL_AudioDeviceID dev;
AudioData ad;
int sampleNum = 0;
};
} // End of namespace AudioGen
int main(int argc, char *argv[])
{
AudioGen::Speaker speaker;
//speaker.pushBeep(440, 1000);
speaker.pushBeep2(440.0f, 1000);
return 0;
}
I've been looking for some time for a quick and reliable way of creating grayscale videos with avconv library from frames that are captured/created with OpenCV, in a C++ application.
I know that OpenCV have it's internal way of creating videos, however, it has some encoding performance and options limitations.
So, in this way, I would like to know which are the options for accomplishing this task?
For accomplishing this task, one of the options that I've found, which complies with my needs is the following class:
extern "C" {
#include <libavcodec/avcodec.h>
#include <libavutil/mathematics.h>
}
class AVConvVideoMaker
{
AVCodec* codec;
AVCodecContext* context;
AVFrame* picture;
int imgWidth, imgHeight, imgBufferSize;
FILE* outputFile;
int outputBufferSize;
uint8_t* outputBuffer;
int pictureBufferSize;
uint8_t* pictureBuffer;
int outputSize;
public:
AVConvVideoMaker(std::string outputFilePath, int imgWidth, int imgHeight)
: codec(NULL)
, context(NULL)
, picture(NULL)
, imgWidth(imgWidth)
, imgHeight(imgHeight)
, imgBufferSize(imgWidth*imgHeight)
, outputFile(fopen(outputFilePath.c_str(), "wb"))
, outputBufferSize(100000)
, outputBuffer(new uint8_t[outputBufferSize])
, pictureBufferSize((imgBufferSize*3)/2)
, pictureBuffer(new uint8_t[pictureBufferSize])
, outputSize(0)
{
avcodec_register_all();
this->codec = avcodec_find_encoder(CODEC_ID_MPEG1VIDEO);
if (!this->codec)
{
throw std::runtime_error("Codec not found");
}
this->context = avcodec_alloc_context3(codec);
this->picture = avcodec_alloc_frame();
this->context->bit_rate = 400000;
this->context->width = this->imgWidth;
this->context->height = this->imgHeight;
this->context->time_base = (AVRational){1, 25};
this->context->gop_size = 10;
this->context->max_b_frames = 1;
this->context->pix_fmt = PIX_FMT_YUV420P;
if(avcodec_open2(this->context, this->codec, NULL) < 0)
{
throw std::runtime_error("Could not open codec");
}
if(!this->outputFile)
{
throw std::runtime_error("Could not open video output file");
}
this->picture->data[0] = this->pictureBuffer;
this->picture->data[1] = this->picture->data[0] + imgBufferSize;
this->picture->data[2] = this->picture->data[1] + imgBufferSize / 4;
this->picture->linesize[0] = this->imgWidth;
this->picture->linesize[1] = this->imgWidth / 2;
this->picture->linesize[2] = this->imgWidth / 2;
}
void insertFrame(cv::Mat1b& img)
{
fflush(stdout);
/* Y */
for(int y=0; y < this->context->height; y++)
{
for(int x=0; x < this->context->width; x++)
{
this->picture->data[0][y * picture->linesize[0] + x] = img.at<uchar>(y,x);
}
}
/* Cb and Cr */
for(int y=0; y < this->context->height/2; y++)
{
for(int x=0; x < this->context->width/2; x++)
{
this->picture->data[1][y * this->picture->linesize[1] + x] = 128;
this->picture->data[2][y * this->picture->linesize[2] + x] = 128;
}
}
this->outputSize = avcodec_encode_video(this->context, this->outputBuffer, this->outputBufferSize, this->picture);
fwrite(this->outputBuffer, 1, outputSize, this->outputFile);
}
~AVConvVideoMaker()
{
this->outputBuffer[0] = 0x00;
this->outputBuffer[1] = 0x00;
this->outputBuffer[2] = 0x01;
this->outputBuffer[3] = 0xb7;
fwrite(this->outputBuffer, 1, 4, this->outputFile);
fclose(this->outputFile);
avcodec_close(this->context);
av_free(this->context);
av_free(this->picture);
delete outputBuffer;
delete pictureBuffer;
}
};
For compiling this in Ubuntu 16.04, you have to link with:
g++ --std=c++11 main.cpp -lopencv_core -lopencv_highgui -lavutil -lavcodec
i have difficulties in using LZMA SDK in my application.
I would like to create a kind of single file compression tool. I dont need any directory support, just need only the LZMA2 stream. But i have no idea on how LZMA SDK is to be used for this.
Please can anyone give me a little example on how the LZMA SDK can be used under C++?
I think that it's a properly little example to use LZMA SDK.
/* LzmaUtil.c -- Test application for LZMA compression
2008-08-05
Igor Pavlov
public domain */
#define _CRT_SECURE_NO_WARNINGS
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "../LzmaDec.h"
#include "../LzmaEnc.h"
#include "../Alloc.h"
const char *kCantReadMessage = "Can not read input file";
const char *kCantWriteMessage = "Can not write output file";
const char *kCantAllocateMessage = "Can not allocate memory";
const char *kDataErrorMessage = "Data error";
static void *SzAlloc(void *p, size_t size) { p = p; return MyAlloc(size); }
static void SzFree(void *p, void *address) { p = p; MyFree(address); }
static ISzAlloc g_Alloc = { SzAlloc, SzFree };
#define kInBufferSize (1 << 15)
#define kOutBufferSize (1 << 15)
unsigned char g_InBuffer[kInBufferSize];
unsigned char g_OutBuffer[kOutBufferSize];
size_t MyReadFile(FILE *file, void *data, size_t size)
{ return fread(data, 1, size, file); }
int MyReadFileAndCheck(FILE *file, void *data, size_t size)
{ return (MyReadFile(file, data, size) == size); }
size_t MyWriteFile(FILE *file, const void *data, size_t size)
{
if (size == 0)
return 0;
return fwrite(data, 1, size, file);
}
int MyWriteFileAndCheck(FILE *file, const void *data, size_t size)
{ return (MyWriteFile(file, data, size) == size); }
long MyGetFileLength(FILE *file)
{
long length;
fseek(file, 0, SEEK_END);
length = ftell(file);
fseek(file, 0, SEEK_SET);
return length;
}
void PrintHelp(char *buffer)
{
strcat(buffer, "\nLZMA Utility 4.58 Copyright (c) 1999-2008 Igor Pavlov 2008-04-11\n"
"\nUsage: lzma <e|d> inputFile outputFile\n"
" e: encode file\n"
" d: decode file\n");
}
int PrintError(char *buffer, const char *message)
{
strcat(buffer, "\nError: ");
strcat(buffer, message);
strcat(buffer, "\n");
return 1;
}
int PrintErrorNumber(char *buffer, SRes val)
{
sprintf(buffer + strlen(buffer), "\nError code: %x\n", (unsigned)val);
return 1;
}
int PrintUserError(char *buffer)
{
return PrintError(buffer, "Incorrect command");
}
#define IN_BUF_SIZE (1 << 16)
#define OUT_BUF_SIZE (1 << 16)
static int Decode(FILE *inFile, FILE *outFile, char *rs)
{
UInt64 unpackSize;
int thereIsSize; /* = 1, if there is uncompressed size in headers */
int i;
int res = 0;
CLzmaDec state;
/* header: 5 bytes of LZMA properties and 8 bytes of uncompressed size */
unsigned char header[LZMA_PROPS_SIZE + 8];
/* Read and parse header */
if (!MyReadFileAndCheck(inFile, header, sizeof(header)))
return PrintError(rs, kCantReadMessage);
unpackSize = 0;
thereIsSize = 0;
for (i = 0; i < 8; i++)
{
unsigned char b = header[LZMA_PROPS_SIZE + i];
if (b != 0xFF)
thereIsSize = 1;
unpackSize += (UInt64)b << (i * 8);
}
LzmaDec_Construct(&state);
res = LzmaDec_Allocate(&state, header, LZMA_PROPS_SIZE, &g_Alloc);
if (res != SZ_OK)
return res;
{
Byte inBuf[IN_BUF_SIZE];
Byte outBuf[OUT_BUF_SIZE];
size_t inPos = 0, inSize = 0, outPos = 0;
LzmaDec_Init(&state);
for (;;)
{
if (inPos == inSize)
{
inSize = MyReadFile(inFile, inBuf, IN_BUF_SIZE);
inPos = 0;
}
{
SizeT inProcessed = inSize - inPos;
SizeT outProcessed = OUT_BUF_SIZE - outPos;
ELzmaFinishMode finishMode = LZMA_FINISH_ANY;
ELzmaStatus status;
if (thereIsSize && outProcessed > unpackSize)
{
outProcessed = (SizeT)unpackSize;
finishMode = LZMA_FINISH_END;
}
res = LzmaDec_DecodeToBuf(&state, outBuf + outPos, &outProcessed,
inBuf + inPos, &inProcessed, finishMode, &status);
inPos += (UInt32)inProcessed;
outPos += outProcessed;
unpackSize -= outProcessed;
if (outFile != 0)
MyWriteFile(outFile, outBuf, outPos);
outPos = 0;
if (res != SZ_OK || thereIsSize && unpackSize == 0)
break;
if (inProcessed == 0 && outProcessed == 0)
{
if (thereIsSize || status != LZMA_STATUS_FINISHED_WITH_MARK)
res = SZ_ERROR_DATA;
break;
}
}
}
}
LzmaDec_Free(&state, &g_Alloc);
return res;
}
typedef struct _CFileSeqInStream
{
ISeqInStream funcTable;
FILE *file;
} CFileSeqInStream;
static SRes MyRead(void *p, void *buf, size_t *size)
{
if (*size == 0)
return SZ_OK;
*size = MyReadFile(((CFileSeqInStream*)p)->file, buf, *size);
/*
if (*size == 0)
return SZE_FAIL;
*/
return SZ_OK;
}
typedef struct _CFileSeqOutStream
{
ISeqOutStream funcTable;
FILE *file;
} CFileSeqOutStream;
static size_t MyWrite(void *pp, const void *buf, size_t size)
{
return MyWriteFile(((CFileSeqOutStream *)pp)->file, buf, size);
}
static SRes Encode(FILE *inFile, FILE *outFile, char *rs)
{
CLzmaEncHandle enc;
SRes res;
CFileSeqInStream inStream;
CFileSeqOutStream outStream;
CLzmaEncProps props;
enc = LzmaEnc_Create(&g_Alloc);
if (enc == 0)
return SZ_ERROR_MEM;
inStream.funcTable.Read = MyRead;
inStream.file = inFile;
outStream.funcTable.Write = MyWrite;
outStream.file = outFile;
LzmaEncProps_Init(&props);
res = LzmaEnc_SetProps(enc, &props);
if (res == SZ_OK)
{
Byte header[LZMA_PROPS_SIZE + 8];
size_t headerSize = LZMA_PROPS_SIZE;
UInt64 fileSize;
int i;
res = LzmaEnc_WriteProperties(enc, header, &headerSize);
fileSize = MyGetFileLength(inFile);
for (i = 0; i < 8; i++)
header[headerSize++] = (Byte)(fileSize >> (8 * i));
if (!MyWriteFileAndCheck(outFile, header, headerSize))
return PrintError(rs, "writing error");
if (res == SZ_OK)
res = LzmaEnc_Encode(enc, &outStream.funcTable, &inStream.funcTable,
NULL, &g_Alloc, &g_Alloc);
}
LzmaEnc_Destroy(enc, &g_Alloc, &g_Alloc);
return res;
}
int main2(int numArgs, const char *args[], char *rs)
{
FILE *inFile = 0;
FILE *outFile = 0;
char c;
int res;
int encodeMode;
if (numArgs == 1)
{
PrintHelp(rs);
return 0;
}
if (numArgs < 3 || numArgs > 4 || strlen(args[1]) != 1)
return PrintUserError(rs);
c = args[1][0];
encodeMode = (c == 'e' || c == 'E');
if (!encodeMode && c != 'd' && c != 'D')
return PrintUserError(rs);
{
size_t t4 = sizeof(UInt32);
size_t t8 = sizeof(UInt64);
if (t4 != 4 || t8 != 8)
return PrintError(rs, "LZMA UTil needs correct UInt32 and UInt64");
}
inFile = fopen(args[2], "rb");
if (inFile == 0)
return PrintError(rs, "Can not open input file");
if (numArgs > 3)
{
outFile = fopen(args[3], "wb+");
if (outFile == 0)
return PrintError(rs, "Can not open output file");
}
else if (encodeMode)
PrintUserError(rs);
if (encodeMode)
{
res = Encode(inFile, outFile, rs);
}
else
{
res = Decode(inFile, outFile, rs);
}
if (outFile != 0)
fclose(outFile);
fclose(inFile);
if (res != SZ_OK)
{
if (res == SZ_ERROR_MEM)
return PrintError(rs, kCantAllocateMessage);
else if (res == SZ_ERROR_DATA)
return PrintError(rs, kDataErrorMessage);
else
return PrintErrorNumber(rs, res);
}
return 0;
}
int MY_CDECL main(int numArgs, const char *args[])
{
char rs[800] = { 0 };
int res = main2(numArgs, args, rs);
printf(rs);
return res;
}
Also you can see it at:
http://read.pudn.com/downloads151/sourcecode/zip/656407/7z460/C/LzmaUtil/LzmaUtil.c__.htm
http://read.pudn.com/downloads157/sourcecode/zip/698262/LZMA/LzmaUtil.c__.htm
I recently found a nice example, written in C++. Credit goes to GH user Treeki who published the original gist:
// note: -D_7ZIP_ST is required when compiling on non-Windows platforms
// g++ -o lzma_sample -std=c++14 -D_7ZIP_ST lzma_sample.cpp LzmaDec.c LzmaEnc.c LzFind.c
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <memory>
#include "LzmaEnc.h"
#include "LzmaDec.h"
static void *_lzmaAlloc(ISzAllocPtr, size_t size) {
return new uint8_t[size];
}
static void _lzmaFree(ISzAllocPtr, void *addr) {
if (!addr)
return;
delete[] reinterpret_cast<uint8_t *>(addr);
}
static ISzAlloc _allocFuncs = {
_lzmaAlloc, _lzmaFree
};
std::unique_ptr<uint8_t[]> lzmaCompress(const uint8_t *input, uint32_t inputSize, uint32_t *outputSize) {
std::unique_ptr<uint8_t[]> result;
// set up properties
CLzmaEncProps props;
LzmaEncProps_Init(&props);
if (inputSize >= (1 << 20))
props.dictSize = 1 << 20; // 1mb dictionary
else
props.dictSize = inputSize; // smaller dictionary = faster!
props.fb = 40;
// prepare space for the encoded properties
SizeT propsSize = 5;
uint8_t propsEncoded[5];
// allocate some space for the compression output
// this is way more than necessary in most cases...
// but better safe than sorry
// (a smarter implementation would use a growing buffer,
// but this requires a bunch of fuckery that is out of
/// scope for this simple example)
SizeT outputSize64 = inputSize * 1.5;
if (outputSize64 < 1024)
outputSize64 = 1024;
auto output = std::make_unique<uint8_t[]>(outputSize64);
int lzmaStatus = LzmaEncode(
output.get(), &outputSize64, input, inputSize,
&props, propsEncoded, &propsSize, 0,
NULL,
&_allocFuncs, &_allocFuncs);
*outputSize = outputSize64 + 13;
if (lzmaStatus == SZ_OK) {
// tricky: we have to generate the LZMA header
// 5 bytes properties + 8 byte uncompressed size
result = std::make_unique<uint8_t[]>(outputSize64 + 13);
uint8_t *resultData = result.get();
memcpy(resultData, propsEncoded, 5);
for (int i = 0; i < 8; i++)
resultData[5 + i] = (inputSize >> (i * 8)) & 0xFF;
memcpy(resultData + 13, output.get(), outputSize64);
}
return result;
}
std::unique_ptr<uint8_t[]> lzmaDecompress(const uint8_t *input, uint32_t inputSize, uint32_t *outputSize) {
if (inputSize < 13)
return NULL; // invalid header!
// extract the size from the header
UInt64 size = 0;
for (int i = 0; i < 8; i++)
size |= (input[5 + i] << (i * 8));
if (size <= (256 * 1024 * 1024)) {
auto blob = std::make_unique<uint8_t[]>(size);
ELzmaStatus lzmaStatus;
SizeT procOutSize = size, procInSize = inputSize - 13;
int status = LzmaDecode(blob.get(), &procOutSize, &input[13], &procInSize, input, 5, LZMA_FINISH_END, &lzmaStatus, &_allocFuncs);
if (status == SZ_OK && procOutSize == size) {
*outputSize = size;
return blob;
}
}
return NULL;
}
void hexdump(const uint8_t *buf, int size) {
int lines = (size + 15) / 16;
for (int i = 0; i < lines; i++) {
printf("%08x | ", i * 16);
int lineMin = i * 16;
int lineMax = lineMin + 16;
int lineCappedMax = (lineMax > size) ? size : lineMax;
for (int j = lineMin; j < lineCappedMax; j++)
printf("%02x ", buf[j]);
for (int j = lineCappedMax; j < lineMax; j++)
printf(" ");
printf("| ");
for (int j = lineMin; j < lineCappedMax; j++) {
if (buf[j] >= 32 && buf[j] <= 127)
printf("%c", buf[j]);
else
printf(".");
}
printf("\n");
}
}
void testIt(const uint8_t *input, int size) {
printf("Test Input:\n");
hexdump(input, size);
uint32_t compressedSize;
auto compressedBlob = lzmaCompress(input, size, &compressedSize);
if (compressedBlob) {
printf("Compressed:\n");
hexdump(compressedBlob.get(), compressedSize);
} else {
printf("Nope, we screwed it\n");
return;
}
// let's try decompressing it now
uint32_t decompressedSize;
auto decompressedBlob = lzmaDecompress(compressedBlob.get(), compressedSize, &decompressedSize);
if (decompressedBlob) {
printf("Decompressed:\n");
hexdump(decompressedBlob.get(), decompressedSize);
} else {
printf("Nope, we screwed it (part 2)\n");
return;
}
printf("----------\n");
}
void testIt(const char *string) {
testIt((const uint8_t *)string, strlen(string));
}
int main(int argc, char **argv) {
testIt("a");
testIt("here is a cool string");
testIt("here's something that should compress pretty well: abcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdef");
return 0;
}
You can refer to this file on how to use lzma2。
https://github.com/Tencent/libpag/blob/aab6391e455193c8ec5b8e2031b495b3fe77b034/test/framework/utils/LzmaUtil.cpp
/////////////////////////////////////////////////////////////////////////////////////////////////
//
// Tencent is pleased to support the open source community by making libpag available.
//
// Copyright (C) 2021 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file
// except in compliance with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// unless required by applicable law or agreed to in writing, software distributed under the
// license is distributed on an "as is" basis, without warranties or conditions of any kind,
// either express or implied. see the license for the specific language governing permissions
// and limitations under the license.
//
/////////////////////////////////////////////////////////////////////////////////////////////////
#include "LzmaUtil.h"
#include "test/framework/lzma/Lzma2DecMt.h"
#include "test/framework/lzma/Lzma2Enc.h"
namespace pag {
static void* LzmaAlloc(ISzAllocPtr, size_t size) {
return new uint8_t[size];
}
static void LzmaFree(ISzAllocPtr, void* address) {
if (!address) {
return;
}
delete[] reinterpret_cast<uint8_t*>(address);
}
static ISzAlloc gAllocFuncs = {LzmaAlloc, LzmaFree};
class SequentialOutStream {
public:
virtual ~SequentialOutStream() = default;
virtual bool write(const void* data, size_t size) = 0;
};
class SequentialInStream {
public:
virtual ~SequentialInStream() = default;
virtual bool read(void* data, size_t size, size_t* processedSize) = 0;
};
struct CSeqInStreamWrap {
ISeqInStream vt;
std::unique_ptr<SequentialInStream> inStream;
};
struct CSeqOutStreamWrap {
ISeqOutStream vt;
std::unique_ptr<SequentialOutStream> outStream;
};
class BuffPtrInStream : public SequentialInStream {
public:
explicit BuffPtrInStream(const uint8_t* buffer, size_t bufferSize)
: buffer(buffer), bufferSize(bufferSize) {
}
bool read(void* data, size_t size, size_t* processedSize) override {
if (processedSize) {
*processedSize = 0;
}
if (size == 0 || position >= bufferSize) {
return true;
}
auto remain = bufferSize - position;
if (remain > size) {
remain = size;
}
memcpy(data, static_cast<const uint8_t*>(buffer) + position, remain);
position += remain;
if (processedSize) {
*processedSize = remain;
}
return true;
}
private:
const uint8_t* buffer = nullptr;
size_t bufferSize = 0;
size_t position = 0;
};
class VectorOutStream : public SequentialOutStream {
public:
explicit VectorOutStream(std::vector<uint8_t>* buffer) : buffer(buffer) {
}
bool write(const void* data, size_t size) override {
auto oldSize = buffer->size();
buffer->resize(oldSize + size);
memcpy(&(*buffer)[oldSize], data, size);
return true;
}
private:
std::vector<uint8_t>* buffer;
};
class BuffPtrSeqOutStream : public SequentialOutStream {
public:
BuffPtrSeqOutStream(uint8_t* buffer, size_t size) : buffer(buffer), bufferSize(size) {
}
bool write(const void* data, size_t size) override {
auto remain = bufferSize - position;
if (remain > size) {
remain = size;
}
if (remain != 0) {
memcpy(buffer + position, data, remain);
position += remain;
}
return remain != 0 || size == 0;
}
private:
uint8_t* buffer = nullptr;
size_t bufferSize = 0;
size_t position = 0;
};
static const size_t kStreamStepSize = 1 << 31;
static SRes MyRead(const ISeqInStream* p, void* data, size_t* size) {
CSeqInStreamWrap* wrap = CONTAINER_FROM_VTBL(p, CSeqInStreamWrap, vt);
auto curSize = (*size < kStreamStepSize) ? *size : kStreamStepSize;
if (!wrap->inStream->read(data, curSize, &curSize)) {
return SZ_ERROR_READ;
}
*size = curSize;
return SZ_OK;
}
static size_t MyWrite(const ISeqOutStream* p, const void* buf, size_t size) {
auto* wrap = CONTAINER_FROM_VTBL(p, CSeqOutStreamWrap, vt);
if (wrap->outStream->write(buf, size)) {
return size;
}
return 0;
}
class Lzma2Encoder {
public:
Lzma2Encoder() {
encoder = Lzma2Enc_Create(&gAllocFuncs, &gAllocFuncs);
}
~Lzma2Encoder() {
Lzma2Enc_Destroy(encoder);
}
std::shared_ptr<Data> code(const std::shared_ptr<Data>& inputData) {
if (encoder == nullptr || inputData == nullptr || inputData->size() == 0) {
return nullptr;
}
auto inputSize = inputData->size();
CLzma2EncProps lzma2Props;
Lzma2EncProps_Init(&lzma2Props);
lzma2Props.lzmaProps.dictSize = inputSize;
lzma2Props.lzmaProps.level = 9;
lzma2Props.numTotalThreads = 4;
Lzma2Enc_SetProps(encoder, &lzma2Props);
std::vector<uint8_t> outBuf;
outBuf.resize(1 + 8);
outBuf[0] = Lzma2Enc_WriteProperties(encoder);
for (int i = 0; i < 8; i++) {
outBuf[1 + i] = static_cast<uint8_t>(inputSize >> (8 * i));
}
CSeqInStreamWrap inWrap = {};
inWrap.vt.Read = MyRead;
inWrap.inStream = std::make_unique<BuffPtrInStream>(
static_cast<const uint8_t*>(inputData->data()), inputSize);
CSeqOutStreamWrap outStream = {};
outStream.vt.Write = MyWrite;
outStream.outStream = std::make_unique<VectorOutStream>(&outBuf);
auto status =
Lzma2Enc_Encode2(encoder, &outStream.vt, nullptr, nullptr, &inWrap.vt, nullptr, 0, nullptr);
if (status != SZ_OK) {
return nullptr;
}
return Data::MakeWithCopy(&outBuf[0], outBuf.size());
}
private:
CLzma2EncHandle encoder = nullptr;
};
std::shared_ptr<Data> LzmaUtil::Compress(const std::shared_ptr<Data>& pixelData) {
Lzma2Encoder encoder;
return encoder.code(pixelData);
}
class Lzma2Decoder {
public:
Lzma2Decoder() {
decoder = Lzma2DecMt_Create(&gAllocFuncs, &gAllocFuncs);
}
~Lzma2Decoder() {
if (decoder) {
Lzma2DecMt_Destroy(decoder);
}
}
std::shared_ptr<Data> code(const std::shared_ptr<Data>& inputData) {
if (decoder == nullptr || inputData == nullptr || inputData->size() == 0) {
return nullptr;
}
auto input = static_cast<const uint8_t*>(inputData->data());
auto inputSize = inputData->size() - 9;
Byte prop = static_cast<const Byte*>(input)[0];
CLzma2DecMtProps props;
Lzma2DecMtProps_Init(&props);
props.inBufSize_ST = inputSize;
props.numThreads = 1;
UInt64 outBufferSize = 0;
for (int i = 0; i < 8; i++) {
outBufferSize |= (input[1 + i] << (i * 8));
}
auto outBuffer = new uint8_t[outBufferSize];
CSeqInStreamWrap inWrap = {};
inWrap.vt.Read = MyRead;
inWrap.inStream = std::make_unique<BuffPtrInStream>(input + 9, inputSize);
CSeqOutStreamWrap outWrap = {};
outWrap.vt.Write = MyWrite;
outWrap.outStream = std::make_unique<BuffPtrSeqOutStream>(outBuffer, outBufferSize);
UInt64 inProcessed = 0;
int isMT = false;
auto res = Lzma2DecMt_Decode(decoder, prop, &props, &outWrap.vt, &outBufferSize, 1, &inWrap.vt,
&inProcessed, &isMT, nullptr);
if (res == SZ_OK && inputSize == inProcessed) {
return Data::MakeAdopted(outBuffer, outBufferSize, Data::DeleteProc);
}
delete[] outBuffer;
return nullptr;
}
private:
CLzma2DecMtHandle decoder = nullptr;
};
std::shared_ptr<Data> LzmaUtil::Decompress(const std::shared_ptr<Data>& data) {
Lzma2Decoder decoder;
return decoder.code(data);
}
} // namespace pag
I have noted that the cascades trained with the program opencv_traincascade does not run with the current version of opencv_performance. I've tried to convert the old performance cpp file to load the new types of cascades, but without success. The code is here:
#include "cv.h"
#include "highgui.h"
#include <cstdio>
#include <cmath>
#include <ctime>
#include <math.h>
#include "opencv2/objdetect/objdetect.hpp"
#include "opencv2/highgui/highgui.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include <iostream>
#include <stdio.h>
#ifndef PATH_MAX
#define PATH_MAX 512
#endif /* PATH_MAX */
/*typedef struct HidCascade {
int size;
int count;
} HidCascade;
*/
typedef struct ObjectPos {
float x;
float y;
float width;
int found; /* for reference */
int neghbors;
} ObjectPos;
using namespace std;
using namespace cv;
int main(int argc, char* argv[]) {
int i, j;
char* classifierdir = NULL;
//char* samplesdir = NULL;
int saveDetected = 1;
double scale_factor = 1.1;
float maxSizeDiff = 1.5F;
float maxPosDiff = 1.1F;
/* number of stages. if <=0 all stages are used */
//int nos = -1, nos0;
int width = 25;
int height = 15;
int rocsize;
FILE* info;
FILE* resultados;
char* infoname;
char fullname[PATH_MAX];
//char detfilename[PATH_MAX];
char* filename;
//char detname[] = "det-";
CascadeClassifier cascade;
double totaltime;
if (!(resultados = fopen("resultados.txt", "w"))) {
printf("Cannot create results file.\n");
exit(-1);
}
infoname = (char*) "";
rocsize = 20;
if (argc == 1) {
printf("Usage: %s\n -data <classifier_directory_name>\n"
" -info <collection_file_name>\n"
" [-maxSizeDiff <max_size_difference = %f>]\n"
" [-maxPosDiff <max_position_difference = %f>]\n"
" [-sf <scale_factor = %f>]\n"
" [-ni]\n"
" [-rs <roc_size = %d>]\n"
" [-w <sample_width = %d>]\n"
" [-h <sample_height = %d>]\n", argv[0], maxSizeDiff,
maxPosDiff, scale_factor, rocsize, width, height);
return 0;
}
for (i = 1; i < argc; i++) {
if (!strcmp(argv[i], "-data")) {
classifierdir = argv[++i];
} else if (!strcmp(argv[i], "-info")) {
infoname = argv[++i];
} else if (!strcmp(argv[i], "-maxSizeDiff")) {
maxSizeDiff = (float) atof(argv[++i]);
} else if (!strcmp(argv[i], "-maxPosDiff")) {
maxPosDiff = (float) atof(argv[++i]);
} else if (!strcmp(argv[i], "-sf")) {
scale_factor = atof(argv[++i]);
} else if (!strcmp(argv[i], "-ni")) {
saveDetected = 0;
} else if (!strcmp(argv[i], "-rs")) {
rocsize = atoi(argv[++i]);
} else if (!strcmp(argv[i], "-w")) {
width = atoi(argv[++i]);
} else if (!strcmp(argv[i], "-h")) {
height = atoi(argv[++i]);
}
}
if (!cascade.load(classifierdir)) {
printf("Unable to load classifier from %s\n", classifierdir);
return 1;
}
strcpy(fullname, infoname);
filename = strrchr(fullname, '\\');
if (filename == NULL) {
filename = strrchr(fullname, '/');
}
if (filename == NULL) {
filename = fullname;
} else {
filename++;
}
info = fopen(infoname, "r");
totaltime = 0.0;
if (info != NULL) {
int x, y, width, height;
Mat img;
int hits, missed, falseAlarms;
int totalHits, totalMissed, totalFalseAlarms;
int found;
float distance;
int refcount;
ObjectPos* ref;
int detcount;
ObjectPos* det;
int error = 0;
int* pos;
int* neg;
pos = (int*) cvAlloc(rocsize * sizeof(*pos));
neg = (int*) cvAlloc(rocsize * sizeof(*neg));
for (i = 0; i < rocsize; i++) {
pos[i] = neg[i] = 0;
}
printf("+================================+======+======+======+\n");
printf("| File Name | Hits |Missed| False|\n");
printf("+================================+======+======+======+\n");
fprintf(resultados,
"+================================+======+======+======+\n");
fprintf(resultados,
"| File Name | Hits |Missed| False|\n");
fprintf(resultados,
"+================================+======+======+======+\n");
//fprintf (resultados, "%d\n",framesCnt);
totalHits = totalMissed = totalFalseAlarms = 0;
while (!feof(info)) {
fscanf(info, "%s %d", filename, &refcount);
img = imread(fullname);
if (!img.data) {
cout << "ow" << endl;
return -1;
}
ref = (ObjectPos*) cvAlloc(refcount * sizeof(*ref));
for (i = 0; i < refcount; i++) {
error = (fscanf(info, "%d %d %d %d", &x, &y, &width, &height)
!= 4);
if (error)
break;
ref[i].x = 0.5F * width + x;
ref[i].y = 0.5F * height + y;
ref[i].width = sqrt(0.5F * (width * width + height * height));
ref[i].found = 0;
ref[i].neghbors = 0; //in the new cascade, where to get the neighbors?
}
vector<Rect> obj_detectados;
Rect retang;
if (!error) {
totaltime -= time(0);
cascade.detectMultiScale(img, obj_detectados, scale_factor, 4, 0
//|CV_HAAR_FIND_BIGGEST_OBJECT
// |CV_HAAR_DO_ROUGH_SEARCH
| CV_HAAR_SCALE_IMAGE, Size(25, 15));
totaltime += time(0);
if (obj_detectados.size() == 0) {
detcount = 0;
} else {
detcount = obj_detectados.size();
}
det = (detcount > 0) ?
((ObjectPos*) cvAlloc(detcount * sizeof(*det))) : NULL;
hits = missed = falseAlarms = 0;
for (vector<Rect>::const_iterator r = obj_detectados.begin();
r != obj_detectados.end(); r++, i++) {
Point r1, r2;
r1.x = (r->x);
r1.y = (r->y);
r2.x = (r->x + r->width);
r2.y = (r->y + r->height);
retang.x = r1.x;
retang.y = r1.y;
retang.width = abs(r2.x - r1.x);
retang.height = abs(r2.y - r1.y);
if (saveDetected) {
rectangle(img, retang, Scalar(0, 0, 255), 3, CV_AA);
}
det[i].x = 0.5F*r->width + r->x;
det[i].y = 0.5F*r->height + r->y;
det[i].width = sqrt(0.5F * (r->width * r->width
+ r->height * r->height));
det[i].neghbors = 1; // i don't know if it will work...
// det[i].neghbors = r.neighbors; --- how to do it in the new version??
found = 0;
for (j = 0; j < refcount; j++) {
distance = sqrtf( (det[i].x - ref[j].x) * (det[i].x - ref[j].x) +
(det[i].y - ref[j].y) * (det[i].y - ref[j].y) );
//cout << distance << endl;
if( (distance < ref[j].width * maxPosDiff) &&
(det[i].width > ref[j].width / maxSizeDiff) &&
(det[i].width < ref[j].width * maxSizeDiff) )
{
ref[j].found = 1;
ref[j].neghbors = MAX( ref[j].neghbors, det[i].neghbors );
found = 1;
}
}
if (!found) {
falseAlarms++;
neg[MIN(det[i].neghbors, rocsize - 1)]++;
//neg[MIN(0, rocsize - 1)]++;
}
}
//imshow("teste", img);
if (saveDetected) {
//strcpy(detfilename, detname);
//strcat(detfilename, filename);
//strcpy(filename, detfilename);
imwrite(fullname, img);
//cvvSaveImage(fullname, img);
}
for (j = 0; j < refcount; j++) {
if (ref[j].found) {
hits++;
//pos[MIN(0, rocsize - 1)]++;
pos[MIN(ref[j].neghbors, rocsize - 1)]++;
} else {
missed++;
}
}
totalHits += hits;
totalMissed += missed;
totalFalseAlarms += falseAlarms;
printf("|%32.64s|%6d|%6d|%6d|\n", filename, hits, missed,
falseAlarms);
//printf("+--------------------------------+------+------+------+\n");
fprintf(resultados, "|%32.64s|%6d|%6d|%6d|\n", filename, hits,
missed, falseAlarms);
//fprintf(resultados,
// "+--------------------------------+------+------+------+\n");
fflush(stdout);
if (det) {
cvFree( &det);
det = NULL;
}
} /* if( !error ) */
//char c = (char) waitKey(10);
// if (c == 27)
// exit(0);
cvFree( &ref);
}
fclose(info);
printf("|%32.32s|%6d|%6d|%6d|\n", "Total", totalHits, totalMissed,
totalFalseAlarms);
fprintf(resultados, "|%32.32s|%6d|%6d|%6d|\n", "Total", totalHits,
totalMissed, totalFalseAlarms);
printf("+================================+======+======+======+\n");
fprintf(resultados,
"+================================+======+======+======+\n");
//printf("Number of stages: %d\n", nos);
//printf("Number of weak classifiers: %d\n", numclassifiers[nos - 1]);
printf("Total time: %f\n", totaltime);
fprintf(resultados, "Total time: %f\n", totaltime);
/* print ROC to stdout */
for (i = rocsize - 1; i > 0; i--) {
pos[i - 1] += pos[i];
neg[i - 1] += neg[i];
}
//fprintf(stderr, "%d\n", nos);
for (i = 0; i < rocsize; i++) {
fprintf(stderr, "\t%d\t%d\t%f\t%f\n", pos[i], neg[i],
((float) pos[i]) / (totalHits + totalMissed),
((float) neg[i]) / (totalHits + totalMissed));
}
cvFree( &pos);
cvFree( &neg);
}
return 0;
}
My doubt is about the det[i].neghbors = r.neighbors; in the old performance.cpp. How I retrieve the neighbors in this new version?
Anyone could help me to convert opencv_performance to run the new cascades from opencv_traincascade?
Many thanks!