LungHistoNet

LungHistoNet is a deep learning project that utilizes Vision Transformers to analyze lung histology images. The goal is to enhance the detection and classification of pulmonary diseases through advanced AI techniques.

Website

Lung Injury Annotation App (Docker)

Desktop annotator backed by Google Drive. The container needs outbound HTTPS to Google APIs and a host X11 display for the Tkinter UI.

Image (canonical tag): amirhosseinebrahimi/lung-injury:latest_v1.7

Keep the Docker Hub repository private. The image embeds a Google service account JSON; anyone with the image can extract it. Rotate credentials on Google Cloud if the image is leaked.

Prerequisites

1. Docker installed (Apple Silicon or Intel build on Mac).

2. Docker Hub account: docker login

3. Pull the image:

   docker pull amirhosseinebrahimi/lung-injury:latest_v1.7

macOS

1. Install Homebrew if needed:

   /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

2. Install socat:

   brew install socat

3. Terminal 1 — leave running:

   socat TCP-LISTEN:6000,reuseaddr,fork UNIX-CLIENT:"$DISPLAY"

4. Terminal 2 — start the app:

   docker run -it --rm -e DISPLAY=host.docker.internal:0 \
     amirhosseinebrahimi/lung-injury:latest_v1.7

   Or with Compose:

   docker compose -f docker-compose.mac.yml run --rm app

Linux (Debian/Ubuntu)

1. Allow Docker to use the host display:

   sudo xhost +local:docker

2. Run the app:

   docker run -it --rm \
     -e DISPLAY="$DISPLAY" \
     -v /tmp/.X11-unix:/tmp/.X11-unix \
     amirhosseinebrahimi/lung-injury:latest_v1.7

   Or with Compose:

   docker compose run --rm app

   Or use the helper script (Linux):

   chmod +x lunginjury
   ./lunginjury

Verify image (no GUI)

Check that Python dependencies load inside the container:

docker run --rm amirhosseinebrahimi/lung-injury:latest_v1.7 \
  python -c "import cv2, plotly, PIL; print('ok')"

Build from source

Place lunginsightcloud-fa31002e7988.json in Application/ (file is gitignored). From the repo root:

docker build -t amirhosseinebrahimi/lung-injury:latest_v1.7 ./Application

Run with the same docker run / docker compose commands as above, using your local tag.

Publish to Docker Hub (maintainers)

GitHub Actions workflow .github/workflows/docker.yml builds and pushes on version tags (v*) or manual workflow_dispatch.

Repository secrets:

Secret	Purpose
GOOGLE_SERVICE_ACCOUNT_JSON	Full JSON body for the service account file baked into the image
DOCKERHUB_USERNAME	Docker Hub username
DOCKERHUB_TOKEN	Docker Hub access token

Manual push example:

docker login
docker build -t amirhosseinebrahimi/lung-injury:latest_v1.7 ./Application
docker push amirhosseinebrahimi/lung-injury:latest_v1.7

Run without Docker

From Application/ with the service account JSON present:

pip install -r requirements.txt
python Application.py