1 of 7

Guess the Blur

(Quickdraw Sketch Guessing Game)

By Andrea & Stik

2 of 7

Goal

Our goal was to create a game that would give both the user and the computer a blurred image. From there, both would have to compete against one another to determine what the image is, based on 5 options that were given. The user is given 3 lives, and the computer is only given 1. If the user fails to guess the image correctly before their 3rd life is up, then they lose.

3 of 7

Demo Videos

3 scenarios of what might happen:

When the user wins
When the computer wins
When both the user and the computer win/tie

4 of 7

Analysis / Conclusions

We used both Neutral Networks and Logistic Regressions to make predictions about our model’s accuracy, but we found that using Neural Networks gave us a higher percentage of accuracy.

We had to find the perfect amount to blur our images to keep the game challenging for both the AI and user.

5 of 7

Challenges & Solutions

A couple challenges we had include:

Uploading datasets (i.e. the Quickdraw sketches) to Google Collab
Blurring the images
Finding a balance for the computer’s ability to guess the blurred image (we didn’t want it to guess it accurately every single time)

Our solutions to these issues were to:

Ask the TA’s for help
Look at previous notebooks
Raised our hands again for the TA’s

6 of 7

Future Work

If we had more time, some things we would like to add to our game include:

Score variables to keep track of the user & computer’s points
Different mode levels (hard, easy, medium) that change the level of the picture’s blurring
A timer
More sketches/images

7 of 7

Links

Code: https://colab.research.google.com/drive/1D77aKLpdKCZ_mSnW089MV9p7jI4hLNCf#scrollTo=bOPU0CuQJdw5

Videos:

https://drive.google.com/file/d/1qkLPMenKHQJJT7J2t1d2heBOJ1kVtRhY/view?resourcekey

https://drive.google.com/file/d/1chsVSzSEVfVRjgOW4CygBotz_-0d9iKk/view?resourcekey

https://drive.google.com/file/d/1DsaeGAVcYEtyQMfqDCf2foR26OOupGQ7/view?resourcekey

Thank you!