MJPEG decode failed: -1 in Docker

[nfs-chris]

I have been trying to get the driver running in Docker but it fails when I launch with additional params, e.g.:
roslaunch azure_kinect_ros_driver driver.launch fps:=30 depth_mode:=NFOV_UNBINNED rgb_point_cloud:=false

results in the following errors:

[error] [t=119] /__w/1/s/extern/Azure-Kinect-Sensor-SDK/src/color/uvc_camerareader.cpp (1300): DecodeMJPEGtoBGRA32(). MJPEG decode failed: -1
[error] [t=120] /__w/1/s/extern/Azure-Kinect-Sensor-SDK/src/sdk/k4a.c (269): capturesync_get_capture(device->capturesync, capture_handle, timeout_in_ms) returned failure in k4a_device_get_capture()

I have tried different combinations of environment variables (DISPLAY etc.) and I have previously managed to get it to run on Foxy with a different Dockerfile, but there the node randomly fails after some time (similar to #249) and the official answer was that 20.04 is not yet supported, so I would appreciate help to get it running properly under Melodic.

I am running this on a Linux machine (Ubuntu 20.04) with an NVIDIA GPU that I am passing in via the NVIDIA Container Toolkit.

Here is my Compose file:

version: "3.9"

services:
  kinect:
    build:
      context: .
      dockerfile: azure_kinect.Dockerfile
    devices:
      - /dev/bus/usb:/dev/bus/usb
      - /dev/dri:/dev/dri
      - /dev/nvidia0:/dev/nvidia0
      - /dev/nvidiactl:/dev/nvidiactl
      - /dev/nvidia-uvm:/dev/nvidia-uvm
    privileged: true
    volumes:
      - /tmp/.X11-unix:/tmp/.X11-unix:rw
      - /usr/lib/nvidia-520:/usr/lib/nvidia-520
    environment:
       - DISPLAY=$DISPLAY
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

and Dockerfile:

FROM ros:melodic

WORKDIR /app
COPY azure_kinect/sources.list /app

ARG DEBIAN_FRONTEND=noninteractive
ENV LC_ALL C.UTF-8
ENV LANG C.UTF-8

RUN apt-get update && apt-get install -y --no-install-recommends \
    curl \
    file \
    dpkg-dev \
    qemu \
    binfmt-support \
    qemu-user-static \
    pkg-config \
    python3 \
    gcc \
    g++ \
    git \
    git-lfs \
    nasm \
    cmake \
    apt-transport-https \
    apt-utils \
    build-essential \ 
    software-properties-common \
    zip \
    unzip \
    python3-catkin-pkg \
    python3-catkin-tools \
    python3-rospkg \
    python3-rosinstall \
    python3-rosinstall-generator \
    python3-rosdep \
    python3-wstool \
    python3-vcstool \
    && rm -rf /var/lib/apt/lists/*

RUN apt-get update && apt-get install -y --no-install-recommends \
    ros-${ROS_DISTRO}-desktop \
    ros-${ROS_DISTRO}-joint-state-publisher \
    ros-${ROS_DISTRO}-catkin \
    ros-${ROS_DISTRO}-image-geometry \
    ros-${ROS_DISTRO}-tf2-geometry-msgs \
    ros-${ROS_DISTRO}-camera-info-manager \
    ros-${ROS_DISTRO}-rgbd-launch \
    && rm -rf /var/lib/apt/lists/*

RUN curl -sSL https://packages.microsoft.com/keys/microsoft.asc | sudo apt-key add -
RUN apt-add-repository https://packages.microsoft.com/ubuntu/18.04/prod
RUN apt-get update && apt-get install -y \
    ninja-build \
    doxygen \
    clang \
    gcc-aarch64-linux-gnu \
    g++-aarch64-linux-gnu \
    && rm -rf /var/lib/apt/lists/*

ENV NVIDIA_VISIBLE_DEVICES ${NVIDIA_VISIBLE_DEVICES:-all}
ENV NVIDIA_DRIVER_CAPABILITIES ${NVIDIA_DRIVER_CAPABILITIES:+$NVIDIA_DRIVER_CAPABILITIES,}graphics
RUN echo "/usr/local/nvidia/lib" >> /etc/ld.so.conf.d/nvidia.conf && \
    echo "/usr/local/nvidia/lib64" >> /etc/ld.so.conf.d/nvidia.conf
# Required for non-glvnd setups.
ENV LD_LIBRARY_PATH /usr/lib/x86_64-linux-gnu:/usr/lib/i386-linux-gnu${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}:/usr/local/nvidia/lib:/usr/local/nvidia/lib64

RUN apt-get update && apt-get install -y \
    libgl1-mesa-dev \
    libsoundio-dev \
    libjpeg-dev \
    libvulkan-dev \
    libx11-dev \
    libxcursor-dev \
    libxinerama-dev \
    libxrandr-dev \
    libusb-1.0-0-dev \
    libssl-dev \
    libudev-dev \
    mesa-common-dev \
    uuid-dev \
    libopencv-dev \
    && rm -rf /var/lib/apt/lists/*

# Azure Kinect ROS depends on Version 1.3.0, we also need the dev version additionally to build
RUN curl -sSL https://packages.microsoft.com/ubuntu/18.04/prod/pool/main/libk/libk4a1.3/libk4a1.3_1.3.0_amd64.deb > /tmp/libk4a1.3_1.3.0_amd64.deb
RUN curl -sSL https://packages.microsoft.com/ubuntu/18.04/prod/pool/main/libk/libk4a1.3-dev/libk4a1.3-dev_1.3.0_amd64.deb > /tmp/libk4a1.3-dev_1.3.0_amd64.deb
RUN curl -sSL https://packages.microsoft.com/ubuntu/18.04/prod/pool/main/k/k4a-tools/k4a-tools_1.3.0_amd64.deb > /tmp/k4a-tools_1.3.0_amd64.deb
RUN echo 'libk4a1.3 libk4a1.3/accepted-eula-hash string 0f5d5c5de396e4fee4c0753a21fee0c1ed726cf0316204edda484f08cb266d76' | debconf-set-selections
RUN echo 'libk4a1.3 libk4a1.3/accept-eula select true' | sudo debconf-set-selections
RUN dpkg -i /tmp/libk4a1.3_1.3.0_amd64.deb
RUN dpkg -i /tmp/libk4a1.3-dev_1.3.0_amd64.deb
RUN dpkg -i /tmp/k4a-tools_1.3.0_amd64.deb

# # Alternative installation method
# RUN echo 'libk4a1.3 libk4a1.3/accepted-eula-hash string 0f5d5c5de396e4fee4c0753a21fee0c1ed726cf0316204edda484f08cb266d76' | debconf-set-selections
# RUN echo 'libk4a1.3 libk4a1.3/accept-eula select true' | sudo debconf-set-selections
# RUN apt-get update && apt-get install -y \
#     libk4a1.3 \
#     libk4a1.3-dev \
#     k4a-tools=1.3.0 \
#     && rm -rf /var/lib/apt/lists/*

RUN mkdir -p /etc/udev/rules.d/
COPY azure_kinect/99-k4a.rules /etc/udev/rules.d/99-k4a.rules
RUN chmod a+rwx /etc/udev/rules.d

COPY azure_kinect/run_ros_azure_kinect.sh /app
RUN ["chmod", "+x", "run_ros_azure_kinect.sh"]
ENTRYPOINT ["./run_ros_azure_kinect.sh"]

as well as the script to start the driver node:

#!/bin/bash

DIR=ros_ws

source /opt/ros/melodic/setup.bash

if [ -d $DIR ]; then
    echo "Starting Kinect..."
    cd ros_ws
    source ./devel/setup.bash
else
    mkdir -p ./ros_ws/src
    cd ./ros_ws/src
    catkin_init_workspace
    git clone https://github.com/microsoft/Azure_Kinect_ROS_Driver -b melodic
    cd ..
    catkin_make --force-cmake
    source ./devel/setup.bash
fi

# This throws errors and terminates after throwing an instance of 'k4a::error'
roslaunch azure_kinect_ros_driver driver.launch fps:=30 depth_mode:=NFOV_UNBINNED rgb_point_cloud:=false

Additionally, it would be great if body tracking could also be included. I am currently unsure if I still need additional dependencies or if setting body_tracking_enabled:=true is enough.

Thank you

[JeffR1992]

I'm having the same issue using ROS2 Humble on Ubuntu 22.04 in Docker. The node crashes after 4-5 hours of running (doing long-run tests) and I get an error saying MJPEG decode failed: -1 when using depth_mode of NFOV_UNBINNED and color_resolution of 2160p at a target framerate of 2160p. Any help from the Kinect team would be appreciated, thanks.

Here's a screenshot of the similar error I'm getting