I am trying to seek the given frame in the video using ffmpeg library. I knew that there is av_seek_frame() function but it was recommended to use avformat_seek_file()instead. Something similar mentioned here.
I know that avformat_seek_file() can't always take you to exact frame you want, but this is ok for me. I just want to jump to the nearest keyframe. So i open the video, find videostream and calling it like this:
avformat_seek_file( formatContext, streamId, 0, frameNumber, frameNumber, AVSEEK_FLAG_FRAME )
It always returns 0, so i understand it as correct finish. However, it doesn't work as it should to. I check byte position like here before and after calling avformat_seek_file(). Actually it changes, but it changes always in the same way whenever i'm trying to put different target frame numbers! I mean that byteposition after this call is always same even with different frameNumber values. Obviously, i'm doing something wrong but i don't know what exactly. I don't know if it does matter but i'm using .h264 files for that. I tried different flags, different files, using timestamps instead of frames, flushing buffers before and after and so on but it doesn't work for me. I will be very grateful if someone could show me what is wrong with it.
I had the same issue, see the code bellow (it works for me):
...
checkPosition(input_files[file_index].ctx);
...
void checkPosition(AVFormatContext *is) {
int stream_index = av_find_default_stream_index(is);
//Convert ts to frame
tm = av_rescale(tm, is->streams[stream_index]->time_base.den, is->streams[stream_index]->time_base.num);
tm /= 1000;
//SEEK
if (avformat_seek_file(is, stream_index, INT64_MIN, tm, INT64_MAX, 0) < 0) {
av_log(NULL, AV_LOG_ERROR, "ERROR av_seek_frame: %u\n", tm);
} else {
av_log(NULL, AV_LOG_ERROR, "SUCCEEDED av_seek_frame: %u newPos:%d\n", tm, is->pb->pos);
avcodec_flush_buffers(is->streams[stream_index]->codec);
}
}
your problem may be related to the fact that your input is raw .h264. try using e.g. mp4box to mux it into a .mp4 file, then load the mp4 file with ffmpeg and try to seek to a keyframe again. e.g.:
mp4box -new -add my_file.h264 my_file.mp4
Related
I tried to do it using OpenCV:
std::string input_name = "<path to file>";
cv::VideoCapture cap(input_name);
if(!cap.isOpened()) printf("Invalid video\n");
However, when I try passing a .txt file, VideoCapture reads it successfully for unknown reasons. When I try to find the number of frames in the video,
if(cap.get(cv::CV_CAP_PROP_FRAME_COUNT) == 0) printf("Invalid video\n");
I get a different number of frames each time I execute. I also tried finding the width and height of a frame in the file,
cv::Mat Frame;
if(!cap.read(Frame)) printf("Cannot read frame\n");
else
{
printf("Frame width is %d\n", Frame.cols);
printf("Frame height is %d\n", Frame.rows);
}
their values are 640 and 400 respectively. Why is this happening? Is there a way to check if a given file is a valid video file preferably using OpenCV?
The only really good way to determine if a file is "valid", is to read its content with an appropriate reader-function that understands that very format. There is no simple, universal marker to say "this is a genuine video file, and can't be something else" - and even if there was, someone could on purpose or by mistake put that in the file so that it looks like a "real video", but isn't when you read the rest of the information. Since there are dozens of different video file formats, it's hard to do something more sensible than "try it, see what happens".
As discussed in another answer, cv::VideoCapture::read() will read the file and return an error indication if the file is not valid. If you want to distinguish between "file doesn't exist", you could do some checking before you try to read it, so you can say "file doesn't exist" as opposed to "sorry, but it seems like that's not a valid video file" - although the latter does apply to "there is no such file" too.
Note that I expect cv::VideoCapture to say isOpened() if the file exists, so for a text-file or PDF, it will happily accept that as "opened". You have to actually try to READ the file to tell if it's a video file. However, the documentation for both ::get and ::read is pretty useless in respect of "What happens on bad input", it almost seems like the functions are designed to only work with valid input.
you can use the method VideoCapture::read(), that will return false if:
the video is corrupt
the string path is pointing to an invalid file
example:
cv::VideoCapture cap(input_name);
auto x = cap.read(im);
I'm not familiar with openCV but you can always check for file extension.
kdt wrote a method for this in this question
https://stackoverflow.com/a/874160/4612406
bool hasEnding (std::string const &fullString, std::string const &ending)
{
if (fullString.length() >= ending.length()) {
return (0 == fullString.compare (fullString.length() - ending.length(), ending.length(), ending));
} else {
return false;
}
}
Just pass the name of the file and the file extension you want like ".mp4". You can change this method to take vector for ending string and check for every string in the vector where you can place .mp4 .avi etc.
I'm currently setting up my output context for creating .avi like this:
avformat_alloc_output_context2(&outContext, NULL, NULL, "out.avi");
if (!outContext)
die("Could not allocate output context");
However, the resulting video quality is very unpleasant. As such, I'd like to be able to fetch the installed codecs on the system and use one of them in avformat_alloc_output_context2. Similar to below:
So I guess my two questions are:
How do I create a list (array) containing the installed codecs (as above)?
How do I use one of them in the output container?
If possible, I'd also like to be able to modify output quality (0%-100%) and open the codec configuration window.
First, make your map with string(or whatever) with AVCodecID, like this :
std::map<string, AVCodecID> _codecList;
_codecList["h264"] = AV_CODEC_ID_H264;
_codecList["mpeg4"] = AV_CODEC_ID_MPEG4;
....
Note that FFmpeg does not provide information that which codec is available in what container so you should validate yourself. but you can reference following link(at least it is officlal) : https://en.wikipedia.org/wiki/Comparison_of_video_container_formats
Next thing to do is that find encoder by name, or AVCodecID in following code :
avcodec_find_encoder_by_name("libx264");
avcodec_find_encoder(AV_CODEC_ID_H264);
Both are return AVCodec* so you can use this when calling avformat_new_stream(), like this :
AVCodecID codec_id = (_codecList.find("h264") != _codecList.end()) ?
_codecList.find("h264") : AV_CODEC_ID_NONE;
if(codec_id == AV_CODEC_ID_NONE)
{
return -1;
}
AVCodec* encoder = avcodec_find_encoder(codec_id);
// or you can just get it from avcodec_find_encoder_by_name("libx264");
AVStream* newStream = avformat_new_stream(avFormatContext, encoder);
Thre are so many things when determining video quality. x264, especially has more options. In this case, you can list it by crf value or bitrate things(you can't use both option). You can determine it with AVCodecContext.
AVCodecContex* codec_ctx = newStream->codec;
codec_ctx->bitrate = 1000000 // 1MB
// codec_ctx->qmin = 18;
// codec_ctx->qmin = 31;
Once you done, open it with avcodec_open2
avcodec_open2(avFormatContext, encoder, NULL);
And Don't forget to close when you release it.
avcodec_close(avFormatContext);
There is much to do when you creating your own output stream. If you have deeper experience with it, i think that this answer will be enough.
But If you don't have much experience with FFmpeg, you can find my full example in here(https://github.com/sorrowhill/FFmpegTutorial)
I have a stl queue of pointers to decoded frames on the heap
(e.g. decode_queue<AVFrame *>)
and I want to take those frames from the stl queue and encode them in WAV format. I tried using the examples that come with ffmpeg they break when I try to change the encoder to pcm_s32le/pcm_s16le. For example for decoding_encoding.c example, I just tried to change some parameters and all of a sudden I am getting a floating point exception.
Is there a something I can do? I am really lost.
UPDATE
In decoding_encoding.c i changed the line:
codec = avcodec_find_encoder(AV_CODEC_ID_MP2);
to
codec = avcodec_find_encoder(AV_CODEC_ID_PCM_S16LE);
this changed ended up making the c->frame_size == 0 So the buffer is never created:
buffer_size = av_samples_get_buffer_size(NULL, c->channels, c->frame_size,
c->sample_fmt, 0);
In that case instead of the line:
frame->nb_samples = c->frame_size;
I changed it to
frame->nb_samples = 10000
and
buffer_size = av_samples_get_buffer_size(NULL, c->channels, 10000,
c->sample_fmt, 0);
just to see what happens(basically substituted c->frame_size for 10000). It compiles but it creates an output file but the player cannot find a stream for it or it is filled with garbage. At this point I am still stuck not sure what to do to get the output.WAV file.
Also, is a RIFF header automatically added or do I have to manually add it in as well?
I try to convert videos for playback on Android using H264, AAC codecs and mp4 container. Video plays normaly with non-system players. But system player shows error "Can't play this video".
I found out that the problem is in moov atom, which is writed in the end of the file.
When I use "-movflags +faststart" ffmeg flag to convert video, it plays normal, but when I try to do that programmatically, it gives no result. I use following code:
av_dict_set( &dict, "movflags", "faststart", 0 );
ret = avformat_write_header( ofmt_ctx, &dict );
This code works fine:
av_dict_set( &dict, "movflags", "faststart", 0 );
ret = avformat_write_header( ofmt_ctx, &dict );
But problem is not solved. I still can't play converted videos on Android devices.
I assume that this answer is very late, but still, for anyone who might still be facing the same issue: this might be caused by the AV_CODEC_FLAG_GLOBAL_HEADER not being set in the audio/video AVCodecContext. A lot of guides show that it needs to be set in the AVFormatContext, but it needs to be set in the AVCodecContext before opening it using avcodec_open2.
if (format_context->oformat->flags & AVFMT_GLOBALHEADER) {
video_codec_context->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
}
avcodec_open2(video_codec_context, video_codec, nullptr);
Maybe the video is not compatible with your android phone? Try to convert with h264 baseline profile.
TL;DR
Set url field of AVFormatContext before avformat_write_header.
Why
I hit same issue today, and I found there is a log when calling av_write_trailer:
Unable to re-open output file for the second pass (faststart)
In the movenc.c implementation, we can see it needs s->url to re-open the file:
avio_flush(s->pb);
ret = s->io_open(s, &read_pb, s->url, AVIO_FLAG_READ, NULL);
if (ret < 0) {
av_log(s, AV_LOG_ERROR, "Unable to re-open %s output file for "
"the second pass (faststart)\n", s->url);
goto end;
}
I am trying to save the decoded image file back as a BMP image using the code in CUDA Decoder project.
if (g_bReadback && g_ReadbackSID)
{
CUresult result = cuMemcpyDtoHAsync(g_bFrameData[active_field], pDecodedFrame[active_field], (nDecodedPitch * nHeight * 3 / 2), g_ReadbackSID);
long padded_size = (nWidth * nHeight * 3 );
CString output_file;
output_file.Format(_T("image/sample_45.BMP"));
SaveBMP(g_bFrameData[active_field],nWidth,nHeight,padded_size,output_file );
if (result != CUDA_SUCCESS)
{
printf("cuMemAllocHost returned %d\n", (int)result);
}
}
But the saved image looks like this
Can anybody help me out here what am i doing wrong .. Thank you.
After investigating further, there were several modifications I made to your approach.
pDecodedFrame is actually in some non-RGB format, I think it is NV12 format which I believe is a particular YUV variant.
pDecodedFrame gets converted to an RGB format on the GPU using a particular CUDA kernel
the target buffer for this conversion will either be a surface provided by OpenGL if g_bUseInterop is specified, or else an ordinary region allocated by the driver API version of cudaMalloc if interop is not specified.
The target buffer mentioned above is pInteropFrame (even in the non-interop case). So to make an example for you, for simplicity I chose to only use the non-interop case, because it's much easier to grab the RGB buffer (pInteropFrame) in that case.
The method here copies pInteropFrame back to the host, after it has been populated with the appropriate RGB image by cudaPostProcessFrame. There is also a routine to save the image as a bitmap file. All of my modifications are delineated with comments that include RMC so search for that if you want to find all the changes/additions I made.
To use, drop this file in the cudaDecodeGL project as a replacement for the videoDecodeGL.cpp source file. Then rebuild the project. Then run the executable normally to display the video. To capture a specific frame, run the executable with the nointerop command-line switch, eg. cudaDecodGL nointerop and the video will not display, but the decode operation and frame capture will take place, and the frame will be saved in a framecap.bmp file. If you want to change the specific frame number that is captured, modify the g_FrameCapSelect = 37; variable to some other number besides 37, and recompile.
Here is the replacement for videoDecodeGL.cpp I used pastebin because SO has a limit on the number of characters that can be entered in a question body.
Note that my approach is independent of whether readback is specified. I would recommend not using readback for this sequence.