feat: add jetson nano docker image

ci/Jenkinsfile.docker

@@ -42,6 +42,14 @@ pipeline {
         sh 'ci/build-docker-images.sh gpu_tensorrt'
       }
     }}
+    stage('Build docker nano image') { steps {
+      catchError(buildResult: 'FAILURE', stageResult: 'FAILURE', catchInterruptions: false) {
+        sh 'echo "DISABLED"'
+        sh 'it does work, it have the same issue as https://github.com/pytorch/pytorch/issues/39680'
+        sh 'exit 1'
+        sh 'ci/build-docker-images.sh nano'
+      }
+    }}
     stage('Update dockerhub readme') {
       when {
        expression { BRANCH_NAME == "master" }

ci/build-docker-images.sh

@@ -18,6 +18,7 @@ TARGETS[gpu]="gpu/default"
 TARGETS[gpu_tf]="gpu/tf"
 TARGETS[gpu_torch]="gpu/torch"
 TARGETS[gpu_tensorrt]="gpu_tensorrt/tensorrt"
+TARGETS[nano]="nano/default"
 
 PR_NUMBER=$(echo $GIT_BRANCH | sed -n '/^PR-/s/PR-//gp')
 if [ "$TAG_NAME" ]; then
@@ -31,6 +32,9 @@ else
     TMP_TAG="trash"
 fi
 
+# Setup binfmt to use qemu for arm docker image
+docker run --rm --privileged multiarch/qemu-user-static --reset -p yes
+
 image_url_prefix="jolibrain/deepdetect"
 
 for name in $NAMES; do

docker/nano.Dockerfile

@@ -0,0 +1,161 @@
+# if built on amd64 arch, install these packages to add arm support to docker
+# apt install -y qemu binfmt-support qemu-user-static
+
+FROM nvcr.io/nvidia/deepstream-l4t:5.1-21.02-base AS build
+
+ARG DEEPDETECT_DEFAULT_MODELS=true
+
+RUN export DEBIAN_FRONTEND=noninteractive && \
+    apt-get update -y && apt-get install -y \
+    wget \
+    curl \
+    cmake \
+    git \
+    ccache \
+    automake \
+    rsync \
+    build-essential \
+    pkg-config \
+    zip \
+    g++ \
+    gcc-7 g++-7 \
+    zlib1g-dev \
+    protobuf-compiler \
+    libprotobuf-dev \
+    libgoogle-glog-dev \
+    libgflags-dev \
+    libeigen3-dev \
+    libopencv-dev \
+    libboost-dev \
+    libboost-filesystem-dev \
+    libboost-thread-dev \
+    libboost-system-dev \
+    libboost-stacktrace-dev \
+    libboost-iostreams-dev \
+    libboost-program-options-dev \
+    libboost-test-dev \
+    libboost-regex-dev \
+    libboost-date-time-dev \
+    libboost-chrono-dev \
+    libssl-dev \
+    libgtest-dev \
+    libcurlpp-dev \
+    libcurl4-openssl-dev \
+    libopenblas-dev \
+    libhdf5-dev \
+    libleveldb-dev \
+    libsnappy-dev \
+    liblmdb-dev \
+    libutfcpp-dev \
+    rapidjson-dev \
+    libmapbox-variant-dev \
+    autoconf \
+    libtool-bin \
+    swig \
+    unzip \
+    libgoogle-perftools-dev \
+    libarchive-dev \
+    bash-completion \
+    schedtool \
+    python-numpy \
+    util-linux
+
+RUN apt clean -y
+
+RUN curl -L https://github.com/Kitware/CMake/releases/download/v3.19.6/cmake-3.19.6-Linux-aarch64.sh -o cmake-install.sh
+RUN chmod +x ./cmake-install.sh
+RUN ./cmake-install.sh --prefix=/usr/local --skip-license
+
+ADD . /opt/deepdetect
+WORKDIR /opt/deepdetect/
+
+ENV CCACHE_DIR=/ccache
+#ENV PATH=/usr/lib/ccache:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+
+# Build Deepdetect
+ENV TERM=xterm
+RUN mkdir build && cd build && \
+    cmake .. \
+        -DBUILD_SPDLOG=ON \
+        -DBUILD_PROTOBUF=OFF \
+        -DUSE_HTTP_SERVER_OATPP=ON \
+        -DUSE_CAFFE=OFF  \
+        -DUSE_TENSORRT=ON  \
+        -DUSE_TENSORRT_OSS=ON  \
+        -DCUDA_ARCH="-gencode arch=compute_62,code=sm_62 -gencode arch=compute_53,code=sm_53" \
+        -DJETSON=ON && make
+
+# Copy libs to /tmp/libs for next build stage
+RUN ./docker/get_libs.sh
+
+# Build final Docker image
+FROM nvcr.io/nvidia/deepstream-l4t:5.1-21.02-base AS runtime
+
+LABEL description="DeepDetect deep learning server & API / jetson nano version"
+LABEL maintainer="[email protected]"
+
+# Install tools and dependencies
+RUN export DEBIAN_FRONTEND=noninteractive && \
+    apt-get update -y && apt-get install -y \
+    wget \
+    curl \
+	libopenblas-base \
+	liblmdb0 \
+	libleveldb1v5 \
+    libboost-regex1.62.0 \
+	libgoogle-glog0v5 \
+	libprotobuf10 \
+    libopencv3.2 \
+	libgflags2.2 \
+	libcurl4 \
+	libcurlpp0 \
+	libhdf5-cpp-100 \
+    libboost-atomic1.65.1 \
+    libboost-chrono1.65.1 \
+    libboost-date-time1.65.1 \
+	libboost-filesystem1.65.1 \
+	libboost-thread1.65.1 \
+	libboost-iostreams1.65.1 \
+    libboost-regex1.65.1 \
+    libboost-stacktrace1.65.1 \
+    libboost-system1.65.1 \
+	libarchive13
+
+# Fix permissions
+RUN ln -sf /dev/stdout /var/log/deepdetect.log && \
+    ln -sf /dev/stderr /var/log/deepdetect.log
+
+RUN useradd -ms /bin/bash dd && \
+    chown -R dd:dd /opt
+USER dd
+
+# Copy Deepdetect binaries from previous step
+COPY --from=build /opt/deepdetect/build/main /opt/deepdetect/build/main
+COPY --from=build /opt/deepdetect/build/oatpp-swagger/src/oatpp-swagger/res/ /opt/deepdetect/build/oatpp-swagger/src/oatpp-swagger/res/
+COPY --from=build --chown=dd /opt/deepdetect/datasets/imagenet/corresp_ilsvrc12.txt /opt/models/ggnet/corresp.txt
+COPY --from=build --chown=dd /opt/deepdetect/datasets/imagenet/corresp_ilsvrc12.txt /opt/models/resnet_50/corresp.txt
+COPY --from=build --chown=dd /opt/deepdetect/templates/caffe/googlenet/*prototxt /opt/models/ggnet/
+COPY --from=build --chown=dd /opt/deepdetect/templates/caffe/resnet_50/*prototxt /opt/models/resnet_50/
+COPY --from=build /tmp/lib/* /opt/deepdetect/build/lib/
+COPY --from=build /opt/deepdetect/templates /opt/deepdetect/build/templates
+
+COPY --from=build /opt/deepdetect/get_models.sh /opt/deepdetect/
+COPY --from=build /opt/deepdetect/docker/check-dede-deps.sh /opt/deepdetect/
+COPY --from=build /opt/deepdetect/docker/start-dede.sh /opt/deepdetect/
+
+# External volume to be mapped, e.g. for models or training data
+WORKDIR /opt/models
+
+USER root
+RUN chown -R dd:dd /opt/models
+
+USER dd
+RUN /opt/deepdetect/get_models.sh
+
+# Ensure all libs are presents
+RUN /opt/deepdetect/check-dede-deps.sh
+
+WORKDIR /opt/deepdetect/build/main
+CMD /opt/deepdetect/start-dede.sh -host 0.0.0.0
+VOLUME ["/data"]
+EXPOSE 8080