Recherche avancée

Sur d’autres sites (6507)

• Linear Attribution Model : What Is It and How Does It Work ?

  16 février 2024, par Erin

  Want a more in-depth way to understand the effectiveness of your marketing campaigns ? Then, the linear attribution model could be the answer.

  Although you can choose from several different attribution models, a linear model is ideal for giving value to every touchpoint along the customer journey. It can help you identify your most effective marketing channels and optimise your campaigns. 

  So, without further ado, let’s explore what a linear attribution model is, when you should use it and how you can get started. 

  What is a linear attribution model ?

  A linear attribution model is a multi-touch method of marketing attribution where equal credit is given to each touchpoint. Every marketing channel used across the entire customer journey gets credit, and each is considered equally important. 

  So, if a potential customer has four interactions before converting, each channel gets 25% of the credit.

  

  Let’s look at how linear attribution works in practice using a hypothetical example of a marketing manager, Sally, who is looking for an alternative to Google Analytics. 

  Sally starts her conversion path by reading a Matomo article comparing Matomo to Google Analytics she finds when searching on Google. A few days later she signs up for a webinar she saw on Matomo’s LinkedIn page. Two weeks later, Sally gets a sign-off from her boss and decides to go ahead with Matomo. She visits the website and starts a free trial by clicking on one of the paid Google Ads. 

  Using a linear attribution model, we credit each of the channels Sally uses (organic traffic, organic social, and paid ads), ensuring no channel is overlooked in our marketing analysis. 

  Are there other types of attribution models ?

  Absolutely. There are several common types of attribution models marketing managers can use to measure the impact of channels in different ways. 

  

  • First interaction : Also called a first-touch attribution model, this method gives all the credit to the first channel in the customer journey. This model is great for optimising the top of your sales funnel.
  • Last interaction : Also called a last-touch attribution model, this approach gives all the credit to the last channel the customer interacts with. It’s a great model for optimising the bottom of your marketing funnel. 
  • Last non-direct interaction : This attribution model excludes direct traffic and credits the previous touchpoint. This is a fantastic alternative to a last-touch attribution model, especially if most customers visit your website before converting. 
  • Time decay attribution model : This model adjusts credit according to the order of the touchpoints. Those nearest the conversion get weighted the highest. 
  • Position-based attribution model : This model allocates 40% of the credit to the first and last touchpoints and splits the remaining 20% evenly between every other interaction.

  Why use a linear attribution model ?

  Marketing attribution is vital if you want to understand which parts of your marketing strategy are working. All of the attribution models described above can help you achieve this to some degree, but there are several reasons to choose a linear attribution model in particular. 

  It uses multi-touch attribution

  Unlike single-touch attribution models like first and last interaction, linear attribution is a multi-touch attribution model that considers every touchpoint. This is vital to get a complete picture of the modern customer journey, where customers interact with companies between 20 and 500 times. 

  Single-touch attribution models can be misleading by giving conversion credit to a single channel, especially if it was the customer’s last use. In our example above, Sally’s last interaction with our brand was through a paid ad, but it was hardly the most important. 

  It’s easy to understand

  Attribution models can be complicated, but linear attribution is easy to understand. Every touchpoint gets the same credit, allowing you to see how your entire marketing function works. This simplicity also makes it easy for marketers to take action. 

  It’s great for identifying effective marketing channels

  Because linear attribution is one of the few models that provides a complete view of the customer journey, it’s easy to identify your most common and influential touchpoints. 

  It accounts for the top and bottom of your funnel, so you can also categorise your marketing channels more effectively and make more informed decisions. For example, PPC ads may be a more common bottom-of-the-full touchpoint and should, therefore, not be used to target broad, top-of-funnel search terms.

  Are there any reasons not to use linear attribution ?

  Linear attribution isn’t perfect. Like all attribution models, it has its weaknesses. Specifically, linear attribution can be too simple, dilute conversion credit and unsuitable for long sales cycles.

  

  It can be too simple

  Linear attribution lacks nuance. It only considers touchpoints while ignoring other factors like brand image and your competitors. This is true for most attribution models, but it’s still important to point it out. 

  It can dilute conversion credit

  In reality, not every touchpoint impacts conversions to the same extent. In the example above, the social media post promoting the webinar may have been the most effective touchpoint, but we have no way of measuring this. 

  The risk with using a linear model is that credit can be underestimated and overestimated — especially if you have a long sales cycle. 

  It’s unsuitable for very long sales cycles

  Speaking of long sales cycles, linear attribution models won’t add much value if your customer journey contains dozens of different touchpoints. Credit will get diluted to the point where analysis becomes impossible, and the model will also struggle to measure the precise ways certain touchpoints impact conversions. 

  Should you use a linear attribution model ?

  A linear attribution model is a great choice for any company with shorter sales cycles or a reasonably straightforward customer journey that uses multiple marketing channels. In these cases, it helps you understand the contribution of each touchpoint and find your best channels. 

  It’s also a practical choice for small businesses and startups that don’t have a team of data scientists on staff or the budget to hire outside help. Because it’s so easy to set up and understand, anyone can start generating insights using this model. 

  How to set up a linear attribution model

  Are you sold on the idea of using a linear attribution model ? Then follow the steps below to get started :

  

  Choose a marketing attribution tool

  Given the market is worth $3.1 billion, you won’t be surprised to learn there are plenty of tools to choose from. But choose carefully. The tool you pick can significantly impact your success with attribution modelling. 

  Take Google Analytics, for instance. While GA4 offers several marketing attribution models for free, including linear attribution, it lacks accuracy due to cookie consent rejection and data sampling. 

  Accurate marketing attribution is included as a feature in Matomo Cloud and is available as a plugin for Matomo On-Premise users. We support a full range of attribution models that use 100% accurate data because we don’t use data sampling, and cookie consent isn’t an issue (with the exception of Germany and the UK). That means you can trust our insights.

  Matomo’s marketing attribution is available out of the box, and we also provide access to raw data, allowing you to develop your custom attribution model. 

  Collect data

  The quality of your marketing attribution also depends on the quality and quantity of your data. It’s why you need to avoid a platform that uses data sampling. 

  This should include :

  • General data from your analytics platform, like pages visited and forms filled
  • Goals and conversions, which we’ll discuss in more detail in the next step
  • Campaign tracking data so you can monitor the behaviour of traffic from different referral channels
  • Behavioural data from features like Heatmaps or Session Recordings

  Set up goals and conversions

  You can’t assign conversion values to customer journey touchpoints if you don’t have conversion goals in place. That’s why the next step of the process is to set up conversion tracking in your web analytics platform. 

  Depending on your type of business and the product you sell, conversions could take one of the following forms :

  • A product purchase
  • Signing up for a webinar
  • Downloading an ebook
  • Filling in a form
  • Starting a free trial

  Setting up these kinds of goals is easy if you use Matomo. 

  Just head to the Goals section of the dashboard, click Manage Goals and then click the green Add A New Goal button. 

  Fill in the screen below, and add a Goal Revenue at the bottom of the page. Doing so will mean Matomo can automatically calculate the value of each touchpoint when using your attribution model. 

  

  If your analytics platform allows it, make sure you also set up Event Tracking, which will allow you to analyse how many users start to take a desired action (like filling in a form) but never complete the task. 

  Try Matomo for Free

  Get the web insights you need, without compromising data accuracy.

  Start for free

  No credit card required

  Test and validate

  As we’ve explained, linear attribution is a great model in some scenarios, but it can fall short if you have a long or complex sales funnel. Even if you’re sure it’s the right model for your company, testing and validating is important. 

  Ideally, your chosen attribution tool should make this process pretty straightforward. For example, Matomo’s Marketing Attribution feature makes comparing and contrasting three different attribution models easy. 

  Here we compare the performance of three attribution models—linear, first-touch, and last-non-direct—in Matomo’s Marketing Attribution dashboard, providing straightforward analysis.

  

  If you think linear attribution accurately reflects the value of your channels, you can start to analyse the insights it generates. If not, then consider using another attribution model.

  Don’t forget to take action from your marketing efforts, either. Linear attribution helps you spot the channels that contribute most to conversions, so allocate more resources to those channels and see if you can improve your conversion rate or boost your ROI. 

  Make the most of marketing attribution with Matomo

  A linear attribution model lets you measure every touchpoint in your customer journey. It’s an easy attribution model to start with and lets you identify and optimise your most effective marketing channels. 

  However, accurate data is essential if you want to benefit the most from marketing attribution data. If your web analytics solution doesn’t play nicely with cookies or uses sampled data, then your linear model isn’t going to tell you the whole story. 

  That’s why over 1 million sites trust Matomo’s privacy-focused web analytics, ensuring accurate data for a comprehensive understanding of customer journeys.

  Now you know what linear attribution modelling is, start employing the model today by signing up for a free 21-day trial, no credit card required. 

  Try Matomo for Free

  21 day free trial. No credit card required.

  
  

  Your email address
  

  Your website address
  

  Your Analytics subdomain will be .matomo.cloud (change)

  Choose your analytics subdomain
  

  .matomo.cloud

  I accept the terms and conditions and data processing agreement (DPA)

  
  Your information will be used to create an account on our cloud service. For more information please consult our privacy policy.
  
  
  

  Start improving your websites now

  

• Ffmpeg input seeking - "Invalid NAL unit size"

  30 janvier 2024, par Dimitris

  I'm trying to use ffmpeg to get a 10-second clip from the middle of a video. The execution time of the command is important, that's why I've decided to use combined input & output seeking (as illustrated here).
The input video file is a CMAF with fragmented MP4, duration of 10 minutes.

  


  I'm testing on a Mac, Ffmpeg version is 6.1.1.

  


  This is the command that I'm using :

  


  ffmpeg -nostdin -y -ss 290 -copyts -start_at_zero -i https://devcdn.flowplayer.com/5f07362e-c358-41d0-857a-c64302a3fcc9/cmaf/17bdb16d-71d1-414c-a291-a028bd45b9ec/playlist_360.m3u8 -ss 300.0 -t 10 -vcodec libwebp -lossless 0 -quality 60 -compression_level 2 -loop 0 -an -sn output.webp


  


  Result : no output file is created.

  


  From what I understand it fails to seek position "290" in the video, probably due to "Invalid NAL unit size" errors.

  


  Here's the output :

  


  ffmpeg version N-106797-g580fb6a8c9-tessus Copyright (c) 2000-2022 the FFmpeg developersbuilt with Apple clang version 11.0.0 (clang-1100.0.33.17)configuration: --cc=/usr/bin/clang --prefix=/opt/ffmpeg --extra-version=tessus --enable-avisynth --enable-fontconfig --enable-gpl --enable-libaom --enable-libass --enable-libbluray --enable-libdav1d --enable-libfreetype --enable-libgsm --enable-libmodplug --enable-libmp3lame --enable-libmysofa --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenh264 --enable-libopenjpeg --enable-libopus --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvmaf --enable-libvo-amrwbenc --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-version3 --pkg-config-flags=--static --disable-ffplaylibavutil      57. 24.101 / 57. 24.101libavcodec     59. 27.100 / 59. 27.100libavformat    59. 23.100 / 59. 23.100libavdevice    59.  6.100 / 59.  6.100libavfilter     8. 37.100 /  8. 37.100libswscale      6.  6.100 /  6.  6.100libswresample   4.  6.100 /  4.  6.100libpostproc    56.  5.100 / 56.  5.100Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'https://devcdn.flowplayer.com/5f07362e-c358-41d0-857a-c64302a3fcc9/cmaf/17bdb16d-71d1-414c-a291-a028bd45b9ec/playlist_360.cmfv':Metadata:major_brand     : isomminor_version   : 1compatible_brands: isomavc1dashcmfccreation_time   : 2024-01-30T07:41:03.000000ZDuration: 00:09:56.54, start: 0.083333, bitrate: 458 kb/sStream #0:0[0x1](und): Video: h264 (High) (avc1 / 0x31637661), yuv420p(tv, bt709, progressive), 640x360 [SAR 1:1 DAR 16:9], 1 kb/s, 24 fps, 24 tbr, 90k tbn (default)Metadata:creation_time   : 2024-01-30T07:41:03.000000Zhandler_name    : ETI ISO Video Media Handlervendor_id       : [0][0][0]ffmpeg version 6.1.1-tessus  https://evermeet.cx/ffmpeg/  Copyright (c) 2000-2023 the FFmpeg developers
  built with Apple clang version 11.0.0 (clang-1100.0.33.17)
  configuration: --cc=/usr/bin/clang --prefix=/opt/ffmpeg --extra-version=tessus --enable-avisynth --enable-fontconfig --enable-gpl --enable-libaom --enable-libass --enable-libbluray --enable-libdav1d --enable-libfreetype --enable-libgsm --enable-libmodplug --enable-libmp3lame --enable-libmysofa --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenh264 --enable-libopenjpeg --enable-libopus --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvmaf --enable-libvo-amrwbenc --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs --enable-libxml2 --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-version3 --pkg-config-flags=--static --disable-ffplay
  libavutil      58. 29.100 / 58. 29.100
  libavcodec     60. 31.102 / 60. 31.102
  libavformat    60. 16.100 / 60. 16.100
  libavdevice    60.  3.100 / 60.  3.100
  libavfilter     9. 12.100 /  9. 12.100
  libswscale      7.  5.100 /  7.  5.100
  libswresample   4. 12.100 /  4. 12.100
  libpostproc    57.  3.100 / 57.  3.100
[hls @ 0x7fc7bb904280] Skip ('#EXT-X-VERSION:6')
[hls @ 0x7fc7bb904280] Opening 'https://devcdn.flowplayer.com/5f07362e-c358-41d0-857a-c64302a3fcc9/cmaf/17bdb16d-71d1-414c-a291-a028bd45b9ec/playlist_360.cmfv' for reading
    Last message repeated 2 times
Input #0, hls, from '[**]/playlist_360.m3u8':
  Duration: 00:09:56.46, start: 0.083333, bitrate: 0 kb/s
  Program 0 
    Metadata:
      variant_bitrate : 0
  Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p(tv, bt709), 640x360 [SAR 1:1 DAR 16:9], 1 kb/s, 24 fps, 24 tbr, 90k tbn (default)
    Metadata:
      variant_bitrate : 0
      compatible_brands: isomavc1dashcmfc
      handler_name    : ETI ISO Video Media Handler
      vendor_id       : [0][0][0][0]
      encoder         : Elemental H.264
      major_brand     : isom
      minor_version   : 1
      creation_time   : 2024-01-30T07:41:03.000000Z
Stream mapping:
  Stream #0:0 -> #0:0 (h264 (native) -> webp (libwebp))
[hls @ 0x7fc7bb904280] Opening 'https://devcdn.flowplayer.com/5f07362e-c358-41d0-857a-c64302a3fcc9/cmaf/17bdb16d-71d1-414c-a291-a028bd45b9ec/playlist_360.cmfv' for reading
    Last message repeated 2 times
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1772342253 > 1534).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1538
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1977545460 > 1481).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1485
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1694403391 > 1582).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1586
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1404850266 > 1661).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1665
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (703351242 > 1680).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1684
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-836978648 > 1751).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1755
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (752797651 > 1867).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1871
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1831058223 > 1833).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 1837
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1238958831 > 2067).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2071
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (435683248 > 2090).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2094
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (2136335178 > 2229).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2233
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1468707300 > 2203).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2207
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (482758774 > 2402).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2406
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1079612217 > 2417).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2421
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-608087491 > 2546).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2550
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-1457748625 > 2527).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2531
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1933919710 > 2734).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2738
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1004643870 > 2803).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2807
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-207765435 > 2988).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2992
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-196888537 > 2306).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2310
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1118966683 > 2620).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2624
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1325583054 > 2715).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2719
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-2003602869 > 2906).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2910
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1666330272 > 3085).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 3089
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (-742329993 > 2593).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2597
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (1326266794 > 2347).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2351
[NULL @ 0x7fc7bb804f40] Invalid NAL unit size (2459776 > 2155).
[NULL @ 0x7fc7bb804f40] missing picture in access unit with size 2159
[https @ 0x7fc7ba022a00] Opening 'https://devcdn.flowplayer.com/5f07362e-c358-41d0-857a-c64302a3fcc9/cmaf/17bdb16d-71d1-414c-a291-a028bd45b9ec/playlist_360.cmfv' for reading
[...]
[vost#0:0/libwebp @ 0x7fc7bbb05780] No filtered frames for output stream, trying to initialize anyway.
Output #0, webp, to 'output.webp':
  Metadata:
    encoder         : Lavf60.16.100
  Stream #0:0(und): Video: webp, yuv420p(progressive), 640x360 [SAR 1:1 DAR 16:9], q=2-31, 200 kb/s, 24 fps, 1k tbn (default)
    Metadata:
      variant_bitrate : 0
      compatible_brands: isomavc1dashcmfc
      handler_name    : ETI ISO Video Media Handler
      vendor_id       : [0][0][0][0]
      creation_time   : 2024-01-30T07:41:03.000000Z
      major_brand     : isom
      minor_version   : 1
      encoder         : Lavc60.31.102 libwebp
[out#0/webp @ 0x7fc7bbb04900] video:0kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown
[out#0/webp @ 0x7fc7bbb04900] Output file is empty, nothing was encoded(check -ss / -t / -frames parameters if used)
frame=    0 fps=0.0 q=0.0 Lsize=       0kB time=N/A bitrate=N/A speed=N/A 


  


  What I've tried so far :

  


  


  1. Downloaded the input file to a local directory and used it as input to ffmpeg - same results.

     


  


  2. Used the mp4 file from the playlist directly as an input to ffmpeg - worked but execution time is very slow

     


  


  3. Emmited the input seeking part (-ss 290 -copyts -start_at_zero) from the command - worked but also very slow in terms of execution time

     


  


  


  Any ideas on why I'm getting "Invalid NAL unit size" and how to make the command work with input seeking ?

  


• Why does every encoded frame's size increase after I had use to set one frame to be key in intel qsv of ffmpeg

  22 avril 2021, par TONY

  I used intel's qsv to encode h264 video in ffmpeg. My av codec context settings is like as below :

  


   m_ctx->width = m_width;
    m_ctx->height = m_height;
    m_ctx->time_base = { 1, (int)fps };
    m_ctx->qmin = 10;
    m_ctx->qmax = 35;
    m_ctx->gop_size = 3000;
    m_ctx->max_b_frames = 0;
    m_ctx->has_b_frames = false;
    m_ctx->refs = 2;
    m_ctx->slices = 0;
    m_ctx->codec_id = m_encoder->id;
    m_ctx->codec_type = AVMEDIA_TYPE_VIDEO;
    m_ctx->pix_fmt = m_h264InputFormat;
    m_ctx->compression_level = 4;
    m_ctx->flags &= ~AV_CODEC_FLAG_CLOSED_GOP;
    AVDictionary *param = nullptr;
    av_dict_set(&param, "idr_interval", "0", 0);
    av_dict_set(&param, "async_depth", "1", 0);
    av_dict_set(&param, "forced_idr", "1", 0);


  


  and in the encoding, I set the AVFrame to be AV_PICTURE_TYPE_I when key frame is needed :

  


    if(key_frame){
        encodeFrame->pict_type = AV_PICTURE_TYPE_I;
    }else{
        encodeFrame->pict_type = AV_PICTURE_TYPE_NONE;
    }
    avcodec_send_frame(m_ctx, encodeFrame);
    avcodec_receive_packet(m_ctx, m_packet);
   std::cerr<<"packet size is "<size<<",is key frame "<code>

  


  The strange phenomenon is that if I had set one frame to AV_PICTURE_TYPE_I, then every encoded frame's size after the key frame would increase. If I change the h264 encoder to x264, then it's ok.

  


  The packet size is as below before I call "encodeFrame->pict_type = AV_PICTURE_TYPE_I" :

  


  packet size is 26839
packet size is 2766
packet size is 2794
packet size is 2193
packet size is 1820
packet size is 2542
packet size is 2024
packet size is 1692
packet size is 2095
packet size is 2550
packet size is 1685
packet size is 1800
packet size is 2276
packet size is 1813
packet size is 2206
packet size is 2745
packet size is 2334
packet size is 2623
packet size is 2055


  


  If I call "encodeFrame->pict_type = AV_PICTURE_TYPE_I", then the packet size is as below :

  


  packet size is 23720,is key frame 1
packet size is 23771,is key frame 0
packet size is 23738,is key frame 0
packet size is 23752,is key frame 0
packet size is 23771,is key frame 0
packet size is 23763,is key frame 0
packet size is 23715,is key frame 0
packet size is 23686,is key frame 0
packet size is 23829,is key frame 0
packet size is 23774,is key frame 0
packet size is 23850,is key frame 0