search

kylelhk / Pictionary

CITS5505 Agile Web Development - Group Project
MIT License
1 stars 0 forks source link

Create a backend AI anti-cheat system (to detect letters and numbers in images) #28

Open wintelestr opened 5 months ago

wintelestr commented 5 months ago

In this project, we are developing an advanced AI system designed to enhance integrity in digital environments by detecting textual content within images. This system is particularly aimed at identifying letters and numbers that may indicate cheating or unauthorized information dissemination within images submitted in online platform.

kylelhk commented 5 months ago

If there is no number in all the drawing prompts, I think detecting letters should be enough. Excluding numbers can also lower the chance of false detection as its easy for a stroke or two to shape like a number. And the function should detect only the letters in the current prompt but not all the 26 letters.

This is a nice to have feature but the gallery should be prioritised. Thanks.

kylelhk commented 4 months ago

Future plan for implementation

  1. Capture the drawing: Convert the canvas drawing to an image format.

  2. Send the image to an AI/ML Service: Use an AI/ML service or API that can analyse the image and detect characters. Services like Google Cloud Vision, AWS Rekognition, or OCR APIs can be useful for this purpose.

  3. Compare detected characters with the word(s) to draw: Extract the characters detected by the AI/ML service and compare them with the characters in the fetched word(s).

Note: Base64 encoding is commonly used to transmit image data, and many AI services (e.g. Google Cloud Vision, AWS Rekognition, etc.) can directly process images provided in Base64 format. These services typically expect the image data to be embedded in a JSON request payload.