backgroundremover test

docker-compose.yml

version: '3.3'
services:
  backgroundremover:
    build:
      context: ./
      dockerfile: Dockerfile
      args:
        - UID=1000
        - GID=1001
    container_name: "backgroundremover"
    user: "1000:1001"
    restart: "no"
    volumes:
      - './images:/app:rw'
    working_dir: /app

Dockerfile

FROM python:3.6-slim

ARG UID
ARG GID

RUN groupadd -g "${GID}" app \
    && useradd --create-home --no-log-init -u "${UID}" -g "${GID}" app

RUN apt-get update && \
    apt-get -y install python3-pip && \
    apt-get -y install ffmpeg imagemagick

USER app

RUN pip install --default-timeout=1000 torch torchvision backgroundremover

WORKDIR /app

ENTRYPOINT ["tail"]
CMD ["-f","/dev/null"]

use.txt

docker exec -it backgroundremover /bin/bash

/home/app/.local/bin/backgroundremover -i input.jpeg -m u2net_human_seg -o output.jpeg

convert output.jpeg -colorspace Gray -trim final.png

hi @benperiton , I'm just curious, is there any reason why in use.txt line 3 you use .jpeg as file format for the output instead of directly using .png and prefer to use convert to pass .jpeg into .png? is it related to resolution? or file size?

I know there is a difference incoming form convert call, because of -colorspace Gray -trim, but not sure if have input file as .jpeg is required before applying those transformations (via command options).

No reason at all, I was just playing around with it and that happened to be what I typed :)