Recherche avancée

Sur d’autres sites (6568)

• Decode video with CUDA nccuvid and ffmpeg [closed]

  25 avril 2013, par Oleksandr Kyrpa

  *strong text*I starting to implement custum video decoder that utilize cuda HW decoder to generate YUV frame for next to encode it.

  How can I fill "CUVIDPICPARAMS" struc ???
  Is it possible ?

  My algorithm are :

  For get video stream packet I'm use ffmpeg-dev libs avcodec, avformat...

  My steps :

  1) Open input file :

  avformat_open_input(&ff_formatContext,in_filename,nullptr,nullptr);

  2) Get video stream property's :

  avformat_find_stream_info(ff_formatContext,nullptr);

  3) Get video stream :

  ff_video_stream=ff_formatContext->streams[i];

  4) Get CUDA device and init it :

  cuDeviceGet(&cu_device,0);
  
  CUcontext cu_vid_ctx;

  5) Init video CUDA decoder and set create params :

  CUVIDDECODECREATEINFO *cu_decoder_info=new CUVIDDECODECREATEINFO;
  
  memset(cu_decoder_info,0,sizeof(CUVIDDECODECREATEINFO));
  
  ...
  
  cuvidCreateDecoder(cu_video_decoder,cu_decoder_info);

  6)Read frame data to AVpacket

  av_read_frame(ff_formatContext,ff_packet);

  AND NOW I NEED decode frame packet on CUDA video decoder, in theoretical are :

  cuvidDecodePicture(pDecoder,&picParams);

  BUT before I need fill CUVIDPICPARAMS

  CUVIDPICPARAMS picParams ;//=new CUVIDPICPARAMS ;
  memset(&picParams, 0, sizeof(CUVIDPICPARAMS)) ;

  HOW CAN I FILL "CUVIDPICPARAMS" struc ???
  
  
  
  typedef struct _CUVIDPICPARAMS
  
  {
  
      int PicWidthInMbs;      // Coded Frame Size
  
      int FrameHeightInMbs;   // Coded Frame Height
  
      int CurrPicIdx;         // Output index of the current picture
  
      int field_pic_flag;     // 0=frame picture, 1=field picture
  
      int bottom_field_flag;  // 0=top field, 1=bottom field (ignored if field_pic_flag=0)
  
      int second_field;       // Second field of a complementary field pair
  
      // Bitstream data
  
      unsigned int nBitstreamDataLen;        // Number of bytes in bitstream data buffer
  
      const unsigned char *pBitstreamData;   // Ptr to bitstream data for this picture (slice-layer)
  
      unsigned int nNumSlices;               // Number of slices in this picture
  
      const unsigned int *pSliceDataOffsets; // nNumSlices entries, contains offset of each slice within the bitstream data buffer
  
      int ref_pic_flag;       // This picture is a reference picture
  
      int intra_pic_flag;     // This picture is entirely intra coded
  
      unsigned int Reserved[30];             // Reserved for future use
  
      // Codec-specific data
  
      union {
  
          CUVIDMPEG2PICPARAMS mpeg2;          // Also used for MPEG-1
  
          CUVIDH264PICPARAMS h264;
  
          CUVIDVC1PICPARAMS vc1;
  
          CUVIDMPEG4PICPARAMS mpeg4;
  
          CUVIDJPEGPICPARAMS jpeg;
  
          unsigned int CodecReserved[1024];
  
      } CodecSpecific;
  
  } CUVIDPICPARAMS;
  
  
  
  typedef struct _CUVIDH264PICPARAMS
  
  {
  
      // SPS
  
      int log2_max_frame_num_minus4;
  
      int pic_order_cnt_type;
  
      int log2_max_pic_order_cnt_lsb_minus4;
  
      int delta_pic_order_always_zero_flag;
  
      int frame_mbs_only_flag;
  
      int direct_8x8_inference_flag;
  
      int num_ref_frames;             // NOTE: shall meet level 4.1 restrictions
  
      unsigned char residual_colour_transform_flag;
  
      unsigned char bit_depth_luma_minus8;    // Must be 0 (only 8-bit supported)
  
      unsigned char bit_depth_chroma_minus8;  // Must be 0 (only 8-bit supported)
  
      unsigned char qpprime_y_zero_transform_bypass_flag;
  
      // PPS
  
      int entropy_coding_mode_flag;
  
      int pic_order_present_flag;
  
      int num_ref_idx_l0_active_minus1;
  
      int num_ref_idx_l1_active_minus1;
  
      int weighted_pred_flag;
  
      int weighted_bipred_idc;
  
      int pic_init_qp_minus26;
  
      int deblocking_filter_control_present_flag;
  
      int redundant_pic_cnt_present_flag;
  
      int transform_8x8_mode_flag;
  
      int MbaffFrameFlag;
  
      int constrained_intra_pred_flag;
  
      int chroma_qp_index_offset;
  
      int second_chroma_qp_index_offset;
  
      int ref_pic_flag;
  
      int frame_num;
  
      int CurrFieldOrderCnt[2];
  
      // DPB
  
      CUVIDH264DPBENTRY dpb[16];          // List of reference frames within the DPB
  
      // Quantization Matrices (raster-order)
  
      unsigned char WeightScale4x4[6][16];
  
      unsigned char WeightScale8x8[2][64];
  
      // FMO/ASO
  
      unsigned char fmo_aso_enable;
  
      unsigned char num_slice_groups_minus1;
  
      unsigned char slice_group_map_type;
  
      signed char pic_init_qs_minus26;
  
      unsigned int slice_group_change_rate_minus1;
  
      union
  
      {
  
          unsigned long long slice_group_map_addr;
  
          const unsigned char *pMb2SliceGroupMap;
  
      } fmo;
  
      unsigned int  Reserved[12];
  
      // SVC/MVC
  
      union
  
      {
  
          CUVIDH264MVCEXT mvcext;
  
          CUVIDH264SVCEXT svcext;
  
      };
  
  } CUVIDH264PICPARAMS;

  How can I fill "CUVIDPICPARAMS" struc ???
  Is it possible ?

• Merging audio at specific point with ffmpeg

  16 octobre 2015, par Deepu

  I was trying to merge audio with ffmpeg with the following command but the audio is merging at starting point.

  ffmpeg -i inputvideo.mp4 -i otheraudio.mp3 -filter_complex "[0:a]atrim=start=0,asetpts=PTS[aud1] ;[1:a]atrim=0:250,asetpts=PTS+100/TB[aud2] ; [aud1][aud2]amix=inputs=2[aout]" -map 0:v -map "[aout]" -crf 24 -pix_fmt yuv420p -c:v libx264 -c:a aac -strict -2 -ac 2 finaloutput.mp4

  with this command it is merging audio streams but not specified point of time.

• How to start MP4 video playback from Android JNI ?

  23 octobre 2015, par Jacob Phillips

  I am seeing answers suggesting FFmpeg is the way to go when manipulating videos natively, but can they be played from a native call ?

  I am specifically looking for a native call because I am actually starting the video on two devices next to each other, and calling mediaPlayer.start() from the application layer tends to introduce a small delay that makes the audio to echo, in a sort of maddening way.

  Speculative thoughts are welcome.