sockheadrps/AiMouseMovement

Training an AI to mimic natural human mouse movement

Mouse Movement dataset gathering and ML Network training

The purpose of this application is to gather data to train an ML model to replicate human-like cursor movement behaviour.

TODO:

1. Media query? disallow data from touch screen devices
2. Env variables etc in go to remove sensitive data in repo
3. Script to pull data from mongo into GAN training set files (python -> json txt file)
4. Train model
5. Generate and test the model's output using pyinput or similar
6. Complete data approval error handling

Data Gathering:

https://mouse.sockhead.software/

The front end is a simple frameworkless tool for data gathering, nothing fancy. Its supported by a backend written in GO.

1. Click yellow square, which will turn green.
2. Move the cursor to the red square, and click again.
3. Repeat as many times as you wish.

Visit the client side data gathering tool

Data Approval:

Datapoints can be approved or removed from the inital DB, and approved data is submitted and stored in an "approved" database, for cleaner data. There is an authorized endpoint that easily allows authorized users to view and vet each data point.

Development

Create a new .env file in the server directory:

/server/.env

MONGO_URI=
DEVELOPMENT=
VALIDATION_USER=
VALIDATION_PWD=
DB=
STAGING_COL=
VALIDATED_COL=