aler9 / gst-darknet Goto Github PK

Hello. Great git. However, there is one question, a very small number of frames on the 2080rtx ti video card. About 15-18 frames per second. However, on the original Darknet repository, the Yolov4 model is 50-70 frames. What is the reason for this? I tried to change the architecture of the video card in the make file. Does not help.

[darknetinfer] frames per second: 6,56
[darknetinfer] frames per second: 17,72
[darknetinfer] frames per second: 18,18
[darknetinfer] frames per second: 18,02
[darknetinfer] frames per second: 18,22
[darknetinfer] frames per second: 18,11
[darknetinfer] frames per second: 18,26
[darknetinfer] frames per second: 18,06
[darknetinfer] frames per second: 18,11
[darknetinfer] frames per second: 18,00
[darknetinfer] frames per second: 18,25
[darknetinfer] frames per second: 18,11
[darknetinfer] frames per second: 18,12
[darknetinfer] frames per second: 18,22
[darknetinfer] frames per second: 18,09
[darknetinfer] frames per second: 18,10
[darknetinfer] frames per second: 17,95
[darknetinfer] frames per second: 18,08
[darknetinfer] frames per second: 17,85
[darknetinfer] frames per second: 17,46
[darknetinfer] frames per second: 17,70
[darknetinfer] frames per second: 17,99
[darknetinfer] frames per second: 18,11
[darknetinfer] frames per second: 17,87
[darknetinfer] frames per second: 17,49
[darknetinfer] frames per second: 17,69
[darknetinfer] frames per second: 17,98
[darknetinfer] frames per second: 18,19
[darknetinfer] frames per second: 17,92
[darknetinfer] frames per second: 18,30
[darknetinfer] frames per second: 18,11

Hi, I got the following pipeline:
time gst-launch-1.0 -e -v filesrc location=test.mp4 ! decodebin ! videoconvert ! darknetinfer config=yolov4.cfg weights=yolov4.weights ! darknetrender labels=coco.names ! darknetprint labels=coco.names ! fakesink

The test.mp4 video has the following specifications:
Duration: 00:00:58.28, start: 0.000000, bitrate: 2625 kb/s
Stream #0:00x1: Video: h264 (High) (avc1 / 0x31637661), yuv420p(progressive), 1920x1080, 2623 kb/s, 25 fps, 25 tbr, 90k tbn (default)

But it take this amount of time to process:
real 1m48.151s
user 2m4.744s
sys 0m1.809s

Is this normal?
I was thinking that it suppose to be equal or less time to process than the original video.

On Nvidia Jetson Nano 2GB with Jetpack 4.6
I try to run:

gst-launch-1.0 \
filesrc location=test.mp4 ! decodebin ! videoconvert \
! darknetinfer config=yolov4.cfg weights=yolov4.weights \
! darknetprint labels=coco.names \
! fakesink

And I'm getting this error message:

Opening in BLOCKING MODE 
NvMMLiteOpen : Block : BlockType = 261 
NVMEDIA: Reading vendor.tegra.display-size : status: 6 
NvMMLiteBlockCreate : Block : BlockType = 261 
WARNING: from element /GstPipeline:pipeline0/GstDecodeBin:decodebin0/GstCapsFilter:capsfilter0: not negotiated
Additional debug info:
gstbasetransform.c(1415): gst_base_transform_reconfigure (): /GstPipeline:pipeline0/GstDecodeBin:decodebin0/GstCapsFilter:capsfilter0:
not negotiated
ERROR: from element /GstPipeline:pipeline0/GstDecodeBin:decodebin0/GstQTDemux:qtdemux0: Internal data stream error.
Additional debug info:
qtdemux.c(6073): gst_qtdemux_loop (): /GstPipeline:pipeline0/GstDecodeBin:decodebin0/GstQTDemux:qtdemux0:
streaming stopped, reason not-negotiated (-4)
ERROR: pipeline doesn't want to preroll.
Setting pipeline to NULL ...
Freeing pipeline ...

It works normally on my Ubuntu Desktop, I don't know where the problem may lie