Recherche avancée

Sur d’autres sites (5380)

• Analytics for the Internet of Things : collecting all your things’ data with Piwik to stay in control ?

  25 novembre 2015, par Matthieu Aubry — About

  At Piwik and Piwik PRO, our mission is to create the leading free and open source analytics platform, and supporting global organisations and communities to keep full control over their data.

  Our broad mission started 8 years ago and we focused at first helping people to liberate their website analytics data, then liberate their mobile app analytics data. But it is clear that there is much more than Web + Mobile : data is everywhere and a lot more data is being generated by software, people and their activities, robots, sensors…

  I’d like to share an interesting article which highlights one of the growing trends of technology : the rise of the Internet Of Things : 6 Ways Analytics And The Internet Of Things Will Transform Business.

  Here is an extract :

  The tech industry is no stranger to change, but the data derived from the IoT is taking disruption to a new level.

  At IBM’s Insight conference last month, Bob Picciano, senior vice president of IBM Analytics, talked about the rise of the “cognitive business”, or an enterprise that engages with analytics to improve its customer relations, business processes, and decision-making capabilities.

  There are dueling predictions over how ubiquitous the Internet of Things will be, but most indicate that the marketplace will host between 50 and 75 billion connected objects by 2020, signaling novel challenges for hardware manufacturing and development. Software engineers, likewise, may need to completely revamp programs to better exploit the influx of data, while innovators need to wrestle with the changes wrought by analytics.

  IBM’s Insight event unfolded in light of this wave of disruption. The lineup of corporate presenters converged on the same message : Analytics is for everyone, and your viability in the marketplace depends on it.

  […]

  IBM’s Insight 2015 conference sounded off on the most important trends in data usage and management. It also served a wake-up call for developers, engineers, and tech leaders. As the Internet of Things alters the landscape of analytics, hardware design needs to change, software development requires novel approaches, and tech management must become more agile in order to realize data’s greatest benefits.

  So far there are 1 million websites using Piwik… but what if there could be 10 or 50 million things (sensors, devices) being measured by Piwik ?

  Together we will be creating the best open source and generic analytics platform, that is engineered to last, and designed to help humanity keep control and gain Freedom.

  We aim for Piwik to be the ideal platform to measure the Internet Of Things.

  We’re still at the beginning of this journey and it will take the best of all of us to get there.

  See you on the way !

  PS : if you’d like to get involved with Piwik, we would be glad to welcome you !

• Audio MP2 encoding and decoding producing half data C++ and FFMPEG

  7 juillet 2016, par hockeyislife

  I have 16 bit PCM data with stereo setup which I grab from a microphone.

  Once I get the data I encode it using the following encoder settings

  AVCodec* audio_codec = avcodec_find_encoder(AV_CODEC_ID_MP2);
  
  
  
  AVCodecContext* audio_codec_ctx = avcodec_alloc_context3(audio_codec);
  
  audio_codec_ctx->bit_rate = 64000;
  
  audio_codec_ctx->channels = 2;
  
  audio_codec_ctx->channel_layout = AV_CH_LAYOUT_STEREO;
  
  audio_codec_ctx->sample_rate = 44100;
  
  audio_codec_ctx->sample_fmt = AV_SAMPLE_FMT_S16;

  When I pass the audio data into the encoder I see that it takes 4608 bytes of data each time and encodes it correctly to MP2 data. PCM data grabbed by the microphone is 88320 bytes and the encoder takes 4608 bytes each time and compresses it.

  If I take each 4608 byte section that has been encoded and pass it through a decoder with the same settings as above but with a decoder.

  AVCodecID audio_codec_id = AV_CODEC_ID_MP2;
  
  AVCodec * audio_decodec = avcodec_find_decoder(audio_codec_id);
  
  
  
  audio_decodecContext = avcodec_alloc_context3(audio_decodec); 
  
  audio_decodecContext->bit_rate = 64000;
  
  audio_decodecContext->channels = 2;
  
  audio_decodecContext->channel_layout = AV_CH_LAYOUT_STEREO;
  
  audio_decodecContext->sample_rate = 44100;
  
  audio_decodecContext->sample_fmt = AV_SAMPLE_FMT_S16;

  The decoding works and is successful but when I look at the data size it is exactly half 2034 of what was encoded. I dont understand why that would be. I would of imagined I would get 4608 considering the encoder and decoder are the same.

  Can anyone shed some light into why this would be happening. Anything I should be setting ?

• ffmpeg send intra and gop over a traditionnal tcp (or udp) server

  27 décembre 2015, par the-owner

  The Problem :
  I want to stream a continuous sequence of bitmaps (a video if stuck).
  I want to send the stream over a QTcpSocket/Server (or a QUdpServer why not, or a traditional tcp/udp server), and with the help of ffmpeg librarie.

  Context and Conditions :
  I code in C++, under msvc2013 win-32, QT-IDE, ffmpeg dev 2.8.3.
  We start with a bitmap which is a raw format, a result from GDI+ that I have gotten.
  For very low client connection, so I will use a light video format compression MPEG-1.
  I already have a TCP-Server (QTcpServer), so my goal is to not rebuild another Server for it ...

  Potential algorithm ? :
  I have thought more on the encoding question than the procole one, assume that the protocol could be good.
  It is the very first time that I do it, I don’t want to enroll into something impossible, and obviously I want your insight about this problem which I think since last month (which I have not found a good answer on internet). So, below the outlines of the algorithm that I have imagined :

  Server Side :

  1. I convert the first BMP to YUV format thanks to sws_scale (), we name it P in the following (ffmpeg - libswscale)
  2. I (MPEG1-)encode the picture P (with avcodec_encode_video2()) I get the datas encoded
  3. I serialize the AVFrame/AVpicture P (maybe we will do this step just once)
  4. In the same time I push the datas into a QByteArray (or a simple byte array ?)
  5. I compress P (maybe with QCompress why not)
  6. I send it to the clients.
  7. If the step 6 goes well : we return to number 1 if the next image is a modulus of k, it will be an Intra otherwise a GOP
  8. Otherwise, I wait ? Or I can take into account the lost frame within the code, and I could jump it ? (Inside these question there are questions of server protocols)

  Client Side (summarized)

  1. We get the datas (new Intra or GOP),
  2. We check the error function on the datas received with respect to the other datas received
  3. If the error function return :
  4. "The Error is OK with this new Intra/GOP received" we start decoding them with ffmpeg, and we display the next image of the livestream.
  5. Otherwise, error is not OK, we take the error into account ... (it is not my main problem here)

  Questions :

  It is a solution that I imagined, If the question of server/client protocol are solved, my algorithm is it viable in order to do a livestream ? Could it work ? Is there a better solution ? (another more simpler algorithm or another librarie than ffmpeg for example or something else ?) a link ? Of course critics on my server protocol are welcome.

  I remind, it is just the outlines.