Todo Notes

• Python is holding any upcoming requests while initializing the dataset. Ideally it runs in parallel
  • Python WebSocket Server
  • Python async requests
  • One new connection for each user between Node - Python
• Performance of group counter is extremely slow
  • The calculation of how many images are in one group is inefficient
• All Nodes Iterations (Frontend) performance improvements
  • updateGroupCount fix
• Fix scissor mode to run not in preview mode
• Adjust max image size based on max displayed counter instead of total images
• Add a popup menu when clicking on the max count to change the max count dynamically

Changelog

Frontend

• Drawing of heatmap optimized to only draw the displayed nodes
• Add a dynamic initial scale based on showing 60% of the images on the initial view
• Create cluster optimizations (creating cluster only for displayed nodes)
• Displayed node optimization (no recalculations if not necessary)
• Added preview mode with delayed switch to normal mode. Triggers on draw-intensive actions
• Added dynamic max image size based on images count in the dataset for better performance
• Fully implemented switch between development and production mode for easier deployment
• Added useful instructions in the README file
• Improved performance by resizing images (normal + preview mode)

Backend

• Improved getNodes performance
• Improved Python server run to no longer require an absolute path to init_json
• Improved the image list generator for easier use
• Added useful instructions in the README file
• Fully implemented switch between development and production mode for easier deployment
• Improved nodemon performance with nodemon.json configuration for file watch definitions
• Improved Python server to use less global variables
• Adjusted image sizes for the new resizing in the frontend

Instructions

Prerequisites

1. MySQL database is running with the following tables available

   1. snapshots

      1. user_id
      2. snapshot_id
      3. snapshot_name
      4. created_at
      5. dataset_id
      6. modified_model
      7. count
      8. groups_count
      9. display_count

   2. user_accounts

      1. user_id
      2. user_name
      3. password

2. Python 3.8+ Interpreter (64-bit) is installed with all requirements

   • Run pip install -r requirements.txt
   • Optional: Create a separate virtual environment (e.g. venv, conda)

3. NodeJS 16+ is installed and available

   • Run npm install to install all required packages

4. Linux OS

5. NVIDIA GPU (optional)

6. Images for at least one dataset

Run commands

Useful Database commands

1. sudo service mysql start
   • Starts the MySQL database (after installation)
2. sudo mysql -u root -p
   • Enters the MySQL terminal as root user on initial setup
3. select user from mysql.user
   • Shows all the available MySQL users
4. create user xyz@localhost identified by 'your password'
   • Creates a user accessible from specified host with a defined password
5. show databases
   • Shows all the available databases which the logged in MySQl user can see
6. create database [database name]
   • Creates the database with specified name
7. use [database name]
   • Switches to use the specified database
8. grant all privileges on [database name].* to xyz@localhost
   • Provides all privileges to the tables of the database for the specified user

Creating the Image Lists

1. Run python /images/make_image_list.py
   • The dataset folder name can be specified inside the file
   • The dataset images have to be located under images/datasetName
     • (.../datasetName/image1.jpg, .../datasetName/image2.png, ...)
   • Supported file endings are .jpg and .png
2. The files will be created under /images/dataset_json

Creating Binary files

Note that the Image List has to be generated beforehand!

1. Run npm run debug_bin
   • Executes the script debug_bin specified in package.json
   • Starts the binBuilder with additional debug information
2. The files will be created under /images/bin
   • File name pattern: datasetName#imageIndex.bin (e.g.: test#1.bin, test#2.bin, ...)

Creating the Initialization JSON (Python Machine Learning)

Note that the Image List has to be generated beforehand!

1. Run python /python_code/generate_dataset_json.py with following parameters
   • -n datasetName
   • -i datasetName
   • -a
   • Example: python generate_dataset_json.py -n xxl_data -i xxl_data -a
2. The files will be created under /images/init_json

Starting the Python Backend Server

1. Ensure that the Database Credentials are correctly set in server.py
2. Ensure that the hostname is correctly set in server.py
3. Run python server.py to run the development mode
4. Run python server.py -prod or python server.py --production to run the production mode

Starting the Node Backend Server

1. Run npm run dev to run the development mode
2. Run npm run prod to run the production mode