Autocrack

Written at hackaTUM 2019 by

• Marko Stapfner
• Christoph Honal
• Ilias Sulgin
• Michael Gigler

Demo

http://18.197.54.251:9200

Dependencies

• docker
• docker-compose
• bash

Deployment

Outgoing ports:

• Mattermost: http://localhost:9200/

• Portainer: http://localhost:9201/

• PgAdmin: http://localhost:9202/

• Image server: http://localhost:9203/

• (Chmod: chmod +x app.sh)

• Build: ./app.sh build

• Deploy: ./app.sh up

• Shutdown ./app.sh down

Mattermost setup for testing localhost:9200

Go to http://localhost:9200:

Credentials:

• Admin-Account:
• Email: admin@example.com
• Password: AdminAdmin$1
• Team-Name: rinderhack

Create channels:

• detection
• alerts

Integrations -> Incoming Webhooks -> Add:

• Title: detection
• Description: detection
• Channel: detection
• Lock to this channel: activated
• Set WEBHOOK_DETECTION in mattermost.env to hash key

Integrations -> Incoming Webhooks -> Add:

• Title: alerts
• Description: alerts
• Channel: alerts
• Lock to this channel: activated
• Set WEBHOOK_ALERTS in mattermost.env to hash key

System Console -> Developer:

• Allow untrusted internal connections to: mattermost-connector

Configuration

Edit core.env to configure thresholds, db.env to configure the database access. Edit azure.env to configure the Azure Accounts

The Images that should be sent by the Camera

Place them in ./persistent-data/mock, the images folder is for the already predicted images.

Microsoft Azure Custom Vision

Visit https://www.customvision.ai/

• Create project
• Create resource
• Create tags:
  • cracked
  • uncracked

Train with some cracked and uncracked images