dusty-nv / jetson-inference

Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.
https://developer.nvidia.com/embedded/twodaystoademo
MIT License
7.83k stars 2.98k forks source link

Background-Net on Jetson Nano (Jetpack 4.6.3) #1678

Open leluro opened 1 year ago

leluro commented 1 year ago

Hi dusty, thank you for your good work! I am currently using a Jetson Nano with the nvmm-disabled branch of this repository and everything is working fine so far. However, I would like to use the background net introduced in later branches. When I am building the repository with the --nvmm-disabled flag, the jetson's camera fails to open when running rosrun jetbot_ros jetbot_camera I get the following error message: [ INFO] [1687165304.273932708]: opening camera device csi://0 @ 1280x720 30.000000fps [gstreamer] initialized gstreamer, version 1.14.5.0 terminate called after throwing an instance of 'std::bad_alloc' what(): std::bad_alloc Can you please help me to fix this problem?

dusty-nv commented 1 year ago

@leluro you should be able to build the master branch with NVMM disabled like this:

cd jetson-inference/build
cmake -DNVMM_ENABLED=off ../
make
sudo make install

Does that help? Are you able to run video-viewer csi://0

rosrun jetbot_ros jetbot_camera

It would seem that you are trying to use an older branch of jetbot_ros, because jetbot_camera was removed in lieu of video_source node in ros_deep_learning package: https://github.com/dusty-nv/ros_deep_learning#video_source-node

Also jetbot_ros as a whole is not actively maintained anymore, and I would recommend using ros_deep_learning or Isaac ROS instead.