This repo contains a dockerfile and src files for orriaga's face-classification routines
Build for Jetson TX2, JetPack 3.2-RC Installs Python3, Tensorflow-gpu, OpenCV3.1 (or OpenCV3.3) and supporting libraries
Prereqs: openhorizon/aarch64-tx2-cudabase: container image (CUDA libs in ubuntu xenial for TX2)
Build time on native Jetson TX2: ~8 hours (better to pull the image)
Image size: 9GB using docker --squash
(raw: 15GB)
docker build --force-rm -f Dockerfile.cv3.1-JP3.2RC -t openhorizon/aarch64-tx2-face-classification-opencv3.1:JetPack3.2-RC .
Container image: https://hub.docker.com/r/openhorizon/aarch64-tx2-face-classification-opencv3.1/
docker pull openhorizon/aarch64-tx2-face-classification-opencv3.1:JetPack3.2-RC
xhost + && docker run -it --rm -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix --cap-add=ALL --ipc=host -v /dev:/dev --privileged openhorizon/aarch64-tx2-face-classification-opencv3.1:JetPack3.2-RC python3 video_emotion_color_demo.py
OR, to run the container, and then a different script
xhost + && docker run -it --rm -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix --cap-add=ALL --ipc=host -v /dev:/dev --privileged openhorizon/aarch64-tx2-face-classification-opencv3.1:JetPack3.2-RC /bin/bash
python3 <python demo file of your choice>
Images and code reference: https://github.com/oarriaga/face_classification