changux/luminoth repository overview

Based on nicolaspence/openshift-luminoth-dev-webserver⁠ work.

⁠Luminoth

Currently, supports object detection, but the team at Tryolabs is aiming for much more. It is built in Python, using TensorFlow and Sonnet.

Luminoth is a tool by TryoLabs

https://github.com/tryolabs/luminoth⁠

Tensorflow is a tool by Google

https://www.tensorflow.org/⁠

⁠Dockerfile

FROM debian:stretch-slim

RUN apt-get update
RUN apt-get install -y python3 python3-dev python3-pip python3-numpy python3-setuptools libopenblas-dev python3-lxml libjpeg-dev ffmpeg

RUN pip3 install --no-cache-dir --upgrade pip numpy setuptools pyasn1-modules
RUN pip install --no-cache-dir tensorflow==1.5
RUN pip install --no-cache-dir luminoth

RUN useradd -m -d /app luminoth

ENV LC_ALL=C.UTF-8
ENV LANG=C.UTF-8

EXPOSE 5000

USER luminoth

WORKDIR /app

RUN /usr/local/bin/lumi checkpoint refresh
RUN /usr/local/bin/lumi checkpoint download fast

CMD /usr/local/bin/lumi server web --host 0.0.0.0 --port 5000 --debug --checkpoint fast

⁠Volumes

To try prediction capability, is recommended map a filesystem folder to docker.

-v HOST_FOLDER:/content

⁠Running

docker run --name luminoth -it -v HOST_FOLDER:/content -p 5000:5000 changux/luminoth:latest

⁠Testing

⁠Images

1. Try with images using the Luminoth web tool http://localhost:5000⁠. Here a (sample)⁠
2. Execute: docker exec -it lumi predict image.png

⁠Videos

1. Copy any mp4 video to your HOST_FOLDER (a small video to begin).

2. Execute: docker exec -it luminoth bash -c "lumi predict /content/video.mp4 --checkpoint=fast --save-media-to=/content"

   Where video.mp4 is the name to your own video.

3. When finish, check the pred_video.mp4 generated file.⁠