Problems with gstreamer after flashing my Jetson TX2 with JetPack3.2

Hi,

After flashing my TX2 with JetPack 3.2, I obtain unexpected results when using gstreamer.

** Before flashing TX2, I used the following command to decode an H264 stream using omxh264dec plugin :
gst-launch-1.0 filesrc location =/media/nvidia/6634-3132/Results/Basketball_gstreamer.264 ! h264parse ! omxh264dec ! filesink location=/media/nvidia/6634-3132/Results/Basketball_gstreamer_lc.yuv, and then when I use the VQMT tool to compare the yuv that I obtain with the original one, so I obtain PSNR=44.5 and SSIM=0.98 as average value for the whole video. But now, after flashing my TX2 and when the following steps I obtain PSNR=19.73 and SSIM=0.59.

Someone can help me to resolve the issues ?

Regards,

I am not sure that flashing TX2 caused the problem, maybe it is the conversion from NV12 format obtained by omxh264dec and I420 format (format of the initial file). But that I am sure that before flashing I have always PSNR=45 and SSIM=0.9 for default properties of encoder.

I obtain same results with :

gst-launch-1.0 filesrc location =/media/nvidia/6634-3132/Results/Basketball_lc.264 ! h264parse ! omxh264dec ! videoparse width=640 height=360 framerate=25/1 format=2 ! videoconvert ! filesink location=/media/nvidia/6634-3132/Results/Basketball_lc_parser_conv.yuv

and

gst-launch-1.0 filesrc location =/media/nvidia/6634-3132/Results/Basketball_lc.264 ! h264parse ! omxh264dec ! videoparse width=640 height=360 framerate=25/1 format=2 ! filesink location=/media/nvidia/6634-3132/Results/Basketball_lc_parser.yuv

Hi Nostalgie,
Do you mean you have the same h264 stream Basketball_gstreamer.264, the decoded YUVs are different on r28.1 and r28.2?

Hi Danell,

Thanks for your reply.

No, that I have is the same original yuv file from which I obtain my Basketball_gstreamer.264, and then after decoded this h264 file I don’t obtain the same results as before flashing my TX2. But I resolved this problem, by replacing videoconv by nvvidconv and now I obtain the same results as before for SSIM and PSNR.

Regards ,