ffmpeg在GPU上生成缩略图

我尝试在GPU上以加速方式生成缩略图，因此编译ffmpeg 3.4.1并运行以下命令：

./ffmpeg  -i ~/Videos/1080.mp4 -vf "hwupload_cuda,thumbnail_cuda=100,scale_cuda=107:60,hwdownload,format=yuv420p" -vframes 1 -y ~/Videos/thumbs/thumb%03d.jpg

但是我得到了错误的输出图片（缩略图的颜色是绿色）：

正确的图像必须为黑色。 如何使用正确的颜色制作第一张图像

日志：-loglevelverbose

ffmpeg version N-89707-g89b84cb Copyright (c) 2000-2018 the FFmpeg developers
  built with gcc 4.8 (Ubuntu 4.8.4-2ubuntu1~14.04.3)
  configuration: --cpu=native --enable-pthreads --extra-version=hd --enable-gnutls --enable-vaapi --enable-vdpau --enable-nonfree --enable-gpl --enable-swscale --enable-libx264 --enable-version3 --enable-nvenc --enable-libnpp --enable-cuda --enable-cuvid --enable-cuda-sdk --enable-filter=scale_cuda --enable-filter=thumbnail_cuda --extra-cflags=-I/usr/local/cuda-8.0/include --extra-ldflags=-L/usr/local/cuda-8.0/lib64
  libavutil      56.  7.100 / 56.  7.100
  libavcodec     58.  9.100 / 58.  9.100
  libavformat    58.  3.100 / 58.  3.100
  libavdevice    58.  0.100 / 58.  0.100
  libavfilter     7. 11.100 /  7. 11.100
  libswscale      5.  0.101 /  5.  0.101
  libswresample   3.  0.101 /  3.  0.101
  libpostproc    55.  0.100 / 55.  0.100
[h264 @ 0x392ed40] Reinit context to 1920x1088, pix_fmt: yuv420p
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '/home/alex/Videos/1080.mp4':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    creation_time   : 2012-05-30T20:01:34.000000Z
  Duration: 00:01:14.21, start: 0.000000, bitrate: 5476 kb/s
    Stream #0:0(und): Video: h264 (High), 1 reference frame (avc1 / 0x31637661), yuv420p(left), 1920x1080 (1920x1088), 5321 kb/s, 29.97 fps, 29.97 tbr, 60k tbn, 59.94 tbc (default)
    Metadata:
      creation_time   : 1970-01-01T00:00:00.000000Z
      handler_name    : VideoHandler
    Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 152 kb/s (default)
    Metadata:
      creation_time   : 2012-05-30T20:01:34.000000Z
      handler_name    : IsoMedia File Produced by Google, 5-11-2011
Stream mapping:
  Stream #0:0 -> #0:0 (h264 (native) -> mjpeg (native))
Press [q] to stop, [?] for help
[h264 @ 0x3a38280] Reinit context to 1920x1088, pix_fmt: yuv420p
[Parsed_thumbnail_cuda_1 @ 0x4e3d8c0] batch size: 100 frames
[graph 0 input from stream 0:0 @ 0x4e8b640] w:1920 h:1080 pixfmt:yuv420p tb:1/60000 fr:30000/1001 sar:0/1 sws_param:flags=2
[auto_scaler_0 @ 0x4e8ca00] w:iw h:ih flags:'bicubic' interl:0
[format @ 0x4e8b540] auto-inserting filter 'auto_scaler_0' between the filter 'Parsed_format_4' and the filter 'format'
[Parsed_scale_cuda_2 @ 0x4e89380] w:1920 h:1080 -> w:107 h:60
[swscaler @ 0x4ebf0c0] deprecated pixel format used, make sure you did set range correctly
[auto_scaler_0 @ 0x4e8ca00] w:107 h:60 fmt:yuv420p sar:0/1 -> w:107 h:60 fmt:yuvj420p sar:0/1 flags:0x4
Output #0, image2, to '/home/alex/Videos/thumbs/thumb%03d.jpg':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    encoder         : Lavf58.3.100
    Stream #0:0(und): Video: mjpeg, 1 reference frame, yuvj420p(pc, left), 107x60, q=2-31, 200 kb/s, 29.97 fps, 29.97 tbn, 29.97 tbc (default)
    Metadata:
      creation_time   : 1970-01-01T00:00:00.000000Z
      handler_name    : VideoHandler
      encoder         : Lavc58.9.100 mjpeg
    Side data:
      cpb: bitrate max/min/avg: 0/0/200000 buffer size: 0 vbv_delay: -1
[Parsed_thumbnail_cuda_1 @ 0x4e3d8c0] frame id #76 (pts_time=2.535867) selected from a set of 100 images
No more output streams to write to, finishing.
frame=    1 fps=0.9 q=3.1 Lsize=N/A time=00:00:02.56 bitrate=N/A speed=2.26x    
video:2kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown
Input file #0 (/home/alex/Videos/1080.mp4):
  Input stream #0:0 (video): 115 packets read (134655 bytes); 101 frames decoded; 
  Input stream #0:1 (audio): 1 packets read (9 bytes); 
  Total: 116 packets (134664 bytes) demuxed
Output file #0 (/home/alex/Videos/thumbs/thumb%03d.jpg):
  Output stream #0:0 (video): 1 frames encoded; 1 packets muxed (1864 bytes); 
  Total: 1 packets (1864 bytes) muxed

更新，正在工作 上传前添加格式=nv12

---

-vf“格式=nv12，hwupload\u cuda，缩略图\u cuda，比例\u cuda=1280:72‌​0，下载，格式=‌​nv12“

解决这种情况的最佳方法是在GPU上解码。它将提高性能，因为CPU的负载将减少。可以使用以下命令执行此操作：

ffmpeg -hwaccel cuvid -c:v h264_cuvid -i video_source \
       -vf "thumbnail_cuda=2,scale_cuda=107:60,hwdownload,format=nv12" \
       -vframes 1 frame.jpg

也最好在解码步骤调整大小，如下所示：

ffmpeg -hwaccel cuvid -c:v h264_cuvid -resize 107x60 -i video_source \
       -vf "thumbnail_cuda=2,hwdownload,format=nv12" \
       -vframes 1 frame.jpg

这些命令显示如何解码h264视频，如果需要解码mpeg2、hevc、vp8、vp9、vc1，则可以从以下列表中选择适当的解码器：

ffmpeg -decoders | grep cuvid
 V..... h264_cuvid           Nvidia CUVID H264 decoder (codec h264)
 V..... hevc_cuvid           Nvidia CUVID HEVC decoder (codec hevc)
 V..... mjpeg_cuvid          Nvidia CUVID MJPEG decoder (codec mjpeg)
 V..... mpeg1_cuvid          Nvidia CUVID MPEG1VIDEO decoder (codec mpeg1video)
 V..... mpeg2_cuvid          Nvidia CUVID MPEG2VIDEO decoder (codec mpeg2video)
 V..... mpeg4_cuvid          Nvidia CUVID MPEG4 decoder (codec mpeg4)
 V..... vc1_cuvid            Nvidia CUVID VC1 decoder (codec vc1)
 V..... vp8_cuvid            Nvidia CUVID VP8 decoder (codec vp8)
 V..... vp9_cuvid            Nvidia CUVID VP9 decoder (codec vp9)

---

共享完整日志。日志中没有什么特别的内容，通常是打印的标准输出。仍然共享它。添加日志到问题中。在hwupload之前添加

format=nv12

。使用GPU加速解码生成缩略图有什么意义？GPU上ffmpeg的总执行时间约为0.5秒，CPU为0.01秒。这是完全没有意义的，除非你想生成成吨的JPEG，也就是成吨的JPEG，否则它也可能毫无意义。JPEG编码仍在CPU上，解码比编码快得多。添加下载操作。CPU总是比GPU的“加速”方法更快。伊万回答得很好。我想知道，你在哪里找到关于

缩略图\u cuda

？我想知道更多关于它的信息，但是谷歌只返回你的答案（在这里和video.stackexchange上），我在官方文档（ffmpeg或nvidia）上找不到任何东西。