.project-metadata.yaml

name: Detecting Defects with Semantic Segmentation
description: >-
  A demonstration of how semantic segmentation can be used to detect defects in manufacturing images.
author: Cloudera Inc.
specification_version: 1.0
prototype_version: 2.0
date: "2022-11-29"

environment_variables:
  KAGGLE_USERNAME:
    default: ""
    description: "Optional Kaggle account username"
  KAGGLE_KEY:
    default: ""
    description: "Optional Kaggle account API token associated with username provided above"

runtimes:
  - editor: Workbench
    kernel: Python 3.9
    edition: Standard

tasks:
  - type: run_session
    name: Install Dependencies
    script: scripts/install_dependencies.py
    cpu: 2
    memory: 4

  - type: run_session
    name: Prepare Data
    script: scripts/prepare_data.py
    cpu: 2
    memory: 4

  - type: run_session
    name: Create model training jobs
    script: scripts/create_train_experiments.py
    cpu: 1
    memory: 2

  - type: start_application
    short_summary: Starting tensorboard
    name: Tensorboard Training Logs
    subdomain: tensorboard
    script: scripts/launch_tensorboard.py
    environment_variables:
      TASK_TYPE: START_APPLICATION
    cpu: 1
    memory: 2

  - type: start_application
    short_summary: Starting streamlit application
    name: Manufacturing Defect Detection
    subdomain: streamlit
    script: scripts/launch_app.py
    environment_variables:
      TASK_TYPE: START_APPLICATION
    cpu: 2
    memory: 4