search

dusty-nv / ros_deep_learning

Deep learning inference nodes for ROS / ROS2 with support for NVIDIA Jetson and TensorRT
898 stars 257 forks source link

failed to find model manifest file 'networks/models.json' #123

Open JsSusenka opened 1 year ago

JsSusenka commented 1 year ago

Greetings. I have a slight problem when trying to use detectnet / imagenet in ros2 foxy container. Video viewer node works just fine, but any other will not work. I am sending log output and l4t version below. Thanks for help in advance!

# R32 (release), REVISION: 7.1, GCID: 29818004, BOARD: t210ref, EABI: aarch64, DATE: Sat Feb 19 17:05:08 UTC 2022
[INFO] [launch]: All log files can be found below /root/.ros/log/2023-06-11-08-33-08-160492-ulet-136
[INFO] [launch]: Default logging verbosity is set to INFO
[INFO] [video_source-1]: process started with pid [139]
[INFO] [imagenet-2]: process started with pid [140]
[INFO] [video_output-3]: process started with pid [141]
[video_source-1] 1686472388.686578 [0] video_sour: using network interface eth0 (udp/192.168.0.165) selected arbitrarily from: eth0, docker0
[video_output-3] 1686472388.706598 [0] video_outp: using network interface eth0 (udp/192.168.0.165) selected arbitrarily from: eth0, docker0
[video_source-1] [INFO] [1686472388.707229664] [video_source]: opening video source: /dev/video0
[video_output-3] [INFO] [1686472388.730642765] [video_output]: opening video output: display://0
[imagenet-2] 1686472388.786331 [0]   imagenet: using network interface eth0 (udp/192.168.0.165) selected arbitrarily from: eth0, docker0
[imagenet-2] [ERROR] [1686472388.815899820] [imagenet]: failed to load imageNet model
[video_output-3] [INFO] [1686472388.855196729] [video_output]: video_output node initialized, waiting for messages
[video_source-1] [gstreamer] initialized gstreamer, version 1.14.5.0
[video_source-1] [gstreamer] gstCamera -- attempting to create device v4l2:///dev/video0
[video_source-1] [gstreamer] gstCamera -- found v4l2 device: Full HD webcam
[video_source-1] [gstreamer] v4l2-proplist, device.path=(string)/dev/video0, udev-probed=(boolean)false, device.api=(string)v4l2, v4l2.device.driver=(string)uvcvideo, v4l2.device.card=(string)"Full\ HD\ webcam", v4l2.device.bus_info=(string)usb-70090000.xusb-2.1, v4l2.device.version=(uint)264701, v4l2.device.capabilities=(uint)2216689665, v4l2.device.device_caps=(uint)69206017;
[video_source-1] [gstreamer] gstCamera -- found 12 caps for v4l2 device /dev/video0
[video_source-1] [gstreamer] [0] video/x-raw, format=(string)YUY2, width=(int)1920, height=(int)1080, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction)5/1;
[video_source-1] [gstreamer] [1] video/x-raw, format=(string)YUY2, width=(int)1280, height=(int)960, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 5/1, 3/1 };
[video_source-1] [gstreamer] [2] video/x-raw, format=(string)YUY2, width=(int)1280, height=(int)720, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 10/1, 5/1, 3/1 };
[video_source-1] [gstreamer] [3] video/x-raw, format=(string)YUY2, width=(int)800, height=(int)600, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 10/1, 5/1 };
[video_source-1] [gstreamer] [4] video/x-raw, format=(string)YUY2, width=(int)640, height=(int)480, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [5] video/x-raw, format=(string)YUY2, width=(int)320, height=(int)240, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [6] image/jpeg, width=(int)1920, height=(int)1080, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [7] image/jpeg, width=(int)1280, height=(int)960, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [8] image/jpeg, width=(int)1280, height=(int)720, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [9] image/jpeg, width=(int)800, height=(int)600, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [10] image/jpeg, width=(int)640, height=(int)480, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] [11] image/jpeg, width=(int)320, height=(int)240, pixel-aspect-ratio=(fraction)1/1, framerate=(fraction){ 30/1, 25/1, 20/1 };
[video_source-1] [gstreamer] gstCamera -- selected device profile:  codec=MJPEG format=unknown width=1280 height=720
[video_source-1] [gstreamer] gstCamera pipeline string:
[video_source-1] [gstreamer] v4l2src device=/dev/video0 do-timestamp=true ! image/jpeg, width=(int)1280, height=(int)720 ! jpegdec name=decoder ! video/x-raw ! appsink name=mysink sync=false
[video_source-1] [gstreamer] gstCamera successfully created device v4l2:///dev/video0
[video_source-1] [video]  created gstCamera from v4l2:///dev/video0
[video_source-1] ------------------------------------------------
[video_source-1] gstCamera video options:
[video_source-1] ------------------------------------------------
[video_source-1]   -- URI: v4l2:///dev/video0
[video_source-1]      - protocol:  v4l2
[video_source-1]      - location:  /dev/video0
[video_source-1]   -- deviceType: default
[video_source-1]   -- ioType:     input
[video_source-1]   -- codec:      MJPEG
[video_source-1]   -- codecType:  cpu
[video_source-1]   -- width:      1280
[video_source-1]   -- height:     720
[video_source-1]   -- frameRate:  30
[video_source-1]   -- numBuffers: 4
[video_source-1]   -- zeroCopy:   true
[video_source-1]   -- flipMethod: none
[video_source-1]   -- loop:       0
[video_source-1] ------------------------------------------------
[video_source-1] [gstreamer] opening gstCamera for streaming, transitioning pipeline to GST_STATE_PLAYING
[video_source-1] [gstreamer] gstreamer changed state from NULL to READY ==> mysink
[video_source-1] [gstreamer] gstreamer changed state from NULL to READY ==> capsfilter1
[video_source-1] [gstreamer] gstreamer changed state from NULL to READY ==> decoder
[video_source-1] [gstreamer] gstreamer changed state from NULL to READY ==> capsfilter0
[video_source-1] [gstreamer] gstreamer changed state from NULL to READY ==> v4l2src0
[video_source-1] [gstreamer] gstreamer changed state from NULL to READY ==> pipeline0
[video_source-1] [gstreamer] gstreamer changed state from READY to PAUSED ==> capsfilter1
[video_source-1] [gstreamer] gstreamer changed state from READY to PAUSED ==> decoder
[video_source-1] [gstreamer] gstreamer changed state from READY to PAUSED ==> capsfilter0
[video_source-1] [gstreamer] gstreamer stream status CREATE ==> src
[imagenet-2] [TRT]    failed to find model manifest file 'networks/models.json'
[imagenet-2] [TRT]    couldn't find built-in classification model 'googlenet'
[INFO] [imagenet-2]: process has finished cleanly [pid 140]
[imagenet-2] 
[video_source-1] [INFO] [1686472389.434509622] [video_source]: allocated CUDA memory for 1280x720 image conversion
dusty-nv commented 1 year ago

@JsSusenka can you try mounting your jetson-inference/data directory into the container at /jetson-inference/data ?

When you start the container, add this flag to your docker run command:

-v /path/to/your/jetson-inference/data:/jetson-inference/data

Where /path/to/your/jetson-inference is the location to where you cloned jetson-inference repo on your device outside of the container. This will not only get your the models.json file, but the models will be downloaded and stored to there, and the TensorRT engines will be built to there. Then when you exit/restart the container, those models will persist and won't have to be redownloaded/rebuilt each time.

Fibo27 commented 1 year ago

Hi @dusty-nv - this is an issue with the containers and I had highlighted this here: https://github.com/dusty-nv/ros_deep_learning/issues/123 The same issue persists when we download the latest containers as well. In addition, this issue: https://github.com/dusty-nv/ros_deep_learning/issues/120, has been carried over to your latest versions. BTW, you are doing fantastic work and I just wanted to highlighted some of these items which probably get carried over during various versions. Thanks

dusty-nv commented 1 year ago

@Fibo27 to address this issue, I've re-organized how the containers are started. Please see the updated documentation here:

https://github.com/dusty-nv/ros_deep_learning#installation https://github.com/dusty-nv/jetson-inference/blob/master/docs/aux-docker.md#ros-support

Since the ros_deep_learning nodes depend on the model directory from jetson-inference/data, you should now run jetson-inference's docker/run.sh --ros=foxy command. This will automatically mount the needed directories/files into the container. The copy of the docker scripts in ros_deep_learning has been removed.

dusty-nv commented 1 year ago

In https://github.com/dusty-nv/jetson-inference/commit/c6602dd46fd9a5fd46934db8933cb54b18665bae I've also made it so that /jetson-inference/data/networks/models.json is built-into the container. So even if these mounts are not setup, it will still be able to download the models. Without the data directory mounted, those models will need to be re-downloaded/re-built after the container exits, but at least it will run even with no mounts.