Automating Google Chrome Dinosaur game

Description

Attempted to automate the Chrome Dinosaur game using Convolutional Neural networks. Used keras, highest score till now ~350. Planning to try more AI/DL approaches on this problem, like CNN+LSTM, genitic algorithms, Deep Q learning etc. Have little knowledge of implementing these techniques. Want to learn along the way. Any help and suggestions are welcome.

Tech Stack

Python, Tensorflow, Keras

URL

Ranter

anon404

828

Comments

0

anon404

828

7y

The input is images. I played the game for a while and stored the images whenever I pressed the spacebar in one folder and whenever I did nothing in a different one.
1

-ANGRY-STUDENT-

13628

7y

Dude, your idea is awesome, but gotta tell you something. At about 400ish it gets very laggy. 2149 is my highscore on your page.
1

-ANGRY-STUDENT-

13628

7y

Fml, why can't I post a pic here...
0

calmyourtities

22274

7y

@gitpull
0

anon404

828

7y

Thanks for the advice @gitpull
2

-ANGRY-STUDENT-

13628

7y

@calmyourtities I am using the web app
1

-ANGRY-STUDENT-

13628

7y

@souravsharan, you are whalecum :)
1

MrEdinLaw

15

7y

Im gonna try making something simple in python and se how far i come. Will give you the code
1

wolfram

3505

7y

I guess rule based automation would work fine in this game .. the game focuses more on timing than strategy...in fact there's hardly any stratgy !
0

anon404

828

7y

@wolfram you are right, a lot of people have used rule based approach before, where no learning is needed at all. Even though it is good testing ground for Deep learning.
0

cachoputa

1806

7y

https://github.com/ukupat/...
0

anon404

828

7y

Good idea @Jop-
0

darkmatter2222

170

7y

Ooo, i want to play with this!

Add Comment