mirror of
https://github.com/guillaumemeyer/watermarks-remover.git
synced 2026-08-22 13:11:57 +02:00
Add an optional harness around THU-BPM/MarkDiffusion (Apache-2.0) for controlled image-watermark experiments and an alternative pixel-removal engine: - markdiffusion_harness.py with watermark / detect / purify subcommands for nine image schemes (Tree-Ring, Ring-ID, ROBIN, WIND, SFW, Gaussian-Shading, GaussMarker, PRC, SEAL); same-scheme/same-model detection only - clean_image.py --remove-pixel diffusion runs the DiffusionPurification regeneration attack (blind regeneration; conservative strength 0.3 default) - setup_markdiffusion.sh bootstrap (PyPI pin 1.0.2; --checkout editable clone at pinned commit), requirements-markdiffusion.txt, Dockerfile.markdiffusion, Makefile targets, and mock-based tests (no torch in CI) - Docs: README section, SKILL.md, removal-matrix.md, vendor-notes.md, references/markdiffusion.md
49 lines
2.1 KiB
Docker
49 lines
2.1 KiB
Docker
# Optional local Docker image for the MarkDiffusion image-watermark harness.
|
|
#
|
|
# Build from the repository root:
|
|
# docker build -f Dockerfile.markdiffusion -t watermarks-remover-markdiffusion .
|
|
#
|
|
# The upstream package is installed from PyPI at build time (pinned in
|
|
# requirements-markdiffusion.txt) and is NOT redistributed by this repository.
|
|
# Upstream is Apache-2.0.
|
|
#
|
|
# Vendored fork hardening:
|
|
# - base image pinned by digest (no moving tag drift)
|
|
# - package version pinned exactly in requirements-markdiffusion.txt
|
|
# - pip itself pinned (no unpinned bootstrap step)
|
|
# - runs as an unprivileged user (a parser bug in a crafted image can no
|
|
# longer write files as root inside the container)
|
|
|
|
# python:3.14-slim linux/amd64 digest.
|
|
FROM python:3.14-slim@sha256:ce40764625a4ff50df3548277632e7f96c4e77fe75fa848aae9885476e7df5a4
|
|
|
|
RUN apt-get update \
|
|
&& apt-get install -y --no-install-recommends \
|
|
libgl1 \
|
|
libglib2.0-0 \
|
|
passwd \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
COPY skills/remove-ai-marks/scripts/requirements-markdiffusion.txt /app/requirements-markdiffusion.txt
|
|
COPY skills/remove-ai-marks/scripts/markdiffusion_harness.py /app/markdiffusion_harness.py
|
|
|
|
# torch is pinned (with everything else) in requirements-markdiffusion.txt; the
|
|
# CPU index keeps the image small. No GPU wheel index inside the image — CUDA
|
|
# users should run setup_markdiffusion.sh on the host instead.
|
|
RUN python3 -m pip install --no-cache-dir "pip==26.2.1" \
|
|
&& python3 -m pip install --no-cache-dir --index-url https://download.pytorch.org/whl/cpu "torch>=2.4,<2.11" \
|
|
&& python3 -m pip install --no-cache-dir -r /app/requirements-markdiffusion.txt
|
|
|
|
# Unprivileged runtime user. The harness only reads input files and writes to
|
|
# stdout/tmp, so nothing under /opt, /app, or the mounted data dir needs root.
|
|
RUN useradd --create-home --uid 10001 --shell /usr/sbin/nologin markdiffusion
|
|
USER markdiffusion
|
|
|
|
ENV HOME=/home/markdiffusion \
|
|
PYTHONUNBUFFERED=1 \
|
|
PYTHONDONTWRITEBYTECODE=1 \
|
|
HF_HOME=/home/markdiffusion/.cache/huggingface
|
|
|
|
WORKDIR /app
|
|
ENTRYPOINT ["python3", "/app/markdiffusion_harness.py"]
|