The Danesfield App is a web application for running the Danesfield algorithms and visualizing results. Danesfield addresses the algorithmic challenges of the IARPA CORE3D program by reconstructing semantically meaningful 3D models of buildings and other man-made structures from satellite imagery.

• Download miniconda with Python 3.7

• docker and docker-compose

1. Update base of the miniconda environment

conda upgrade --all

2. Create, activate new danesfield virtual environment

chmod +x conda/install.sh && conda/install.sh

2. Pulling Danesfield docker image from Docker Hub. This image is being maintained by @shiroinekotfs

docker pull shiroinekotfs/danesfield:linux-amd64

3. Install girder and its dependencies

pip install -e server

4. (Optional) You can edit the logon and password of admin in server/init_girder.py

11 ADMIN_USER = os.getenv("GIRDER_ADMIN_USER", "admin") # Logon is admin
12 ADMIN_PASS = os.getenv("GIRDER_ADMIN_PASS", "1234") # Password is 1234

Girder requires that MongoDB and RabbitMQ are running. If you've installed docker-compose, this can be done easily:

docker-compose up -d

You can stop by using this command

docker-compose stop

Run the following commands separately on the machine you wish to host the application on (within the conda environment):

girder serve --host 0.0.0.0 \
python -m girder_worker -l info --concurrency 4

Models used:

1. UNet Semantic Segmentation
2. Building Segmentation
3. Material Classification
4. Roof Geon Extraction
5. Run Metrics

Upload all models to the models folder within the core3d collection, creating it if it doesn't exist.

1. Install @vue/cli on global

npm install -g @vue/cli

2. Create girder assetstore

3. Enable danesfield-server plugin

4. Copy content from .env file and create .env.local and update the content properly

5. Update and run the server

npm install
npm run serve

npm run build
cp client/dist ./girder/clients/web/static/core3d