JetsonHacks car video detection example needs update

Youtube video on installing jetson 3.0 and at the end it shows running a car detection example.
In version 3.2 the command lines for testing are not in the readme and they have changed.

So if you started by running the latest version 3.2 opened a terminal window and typed it in
sudo ./
changed directory to tegra_multimedia_api/samples/backend/

this line no longer works and delivers an error message. It needs to be modified.
./backend 1 …/…/data/Video/sample_outdoor_car_1080p_10fps.h264 H264
–gie-deployfile …/…/data/Model/GoogleNet_one_class/GoogleNet_modified_oneClass_halfHD.prototxt
–gie-modelfile …/…/data/Model/GoogleNet_one_class/GoogleNet_modified_oneClass_halfHD.caffemodel
–gie-forcefp32 0 --gie-proc-interval 1 -fps 10
./backend 1 …/…/data/Video/sample_outdoor_car_1080p_10fps.h264 H264 --trt-deployfile …/…/data/Model/GoogleNet_one_class/GoogleNet_modified_oneClass_halfHD.prototxt --trt-modelfile …/…/data/Model/GoogleNet_one_class/GoogleNet_modified_oneClass_halfHD.caffemodel --trt-forcefp32 0 --trt-proc-interval 1 -fps 10

JetPack 3.2 is a Developers Preview, which typically does not have all the support material correctly in place. The issue in the README is something typical of the DP. Please note that YouTube videos, unless released by NVIDIA, are not “official” and may not reflect current hardware/software combinations.

As always, you can report issues on the JetPack installer here:

@chicagobob123, glad you fix the command. Hey, it works!

It took a while to pop up the video though (I forgot to run… I almost gave up, thinking something went wrong. Wow, really cool.

By the way, do you know how to exit? Ctrl+C seems won’t stop the show. I ended up pressing the reset button.

P.S. (for my own reminder)

  1. can be found under /home/ubuntu (normally, your home is at /home/nvidia ).
  2. the video stops itself. no need to kill it unless you are in hurry.

Thanks for the update of the car video detection --gie to --trt ( I installed 3.1 ) the --gie didnt work until I ran into your post.

With Jetpack 3.3 this example doesn’t run at all.

nvidia@tegra-ubuntu:~/tegra_multimedia_api/samples/backend$ ./backend 1 …/…/data/Video/sample_outdoor_car_1080p_10fps.h264 H264 --trt-deployfile …/…/data/Model/GoogleNet_one_class/GoogleNet_modified_oneClass_halfHD.prototxt --trt-modelfile …/…/data/Model/GoogleNet_one_class/GoogleNet_modified_oneClass_halfHD.caffemodel --trt-forcefp32 0 --trt-proc-interval 1 -fps 10
Net has batch_size, channel, net_height, net_width:1 3 540 960
forced_fp32 has been set to 0(using fp16)
outputs coverage
outputs bboxes



It suddenly started working after I closed the web browser.

I am able to run this sample with ubuntu 16.04 Host and Jetson TX2 with installation. However to run backend sample . I have to omit the option --trt-mode. Anyone knows why?


Running R28.2.1 on a Jetson TX2, I’m able to run the backend sample.

My problem is that the video shows up but not the bounding box supposed to show cars.

I’m supposed to create my own object detection project but if samples not work, not sure it will…

If somebody know how to fix this problem I’ll be very thankful.

I didn’t done the first install (it’s my university board) but I found that it used :
Jetpack 3.0 (supposedly)
CUDA 9.0
TensorRT libs are 4.1.3
OpenCV 3.3.1

Others sample are working well (once I used “export DISPLAY=:0”)