Here I've used deep q network that takes a preprocessed frame as an input and produces the most optimal choice for agent's action.
To train the network : python3 dino.py -m "Train"
To observe : python3 dino.py -m "Run"
Read the full article on medium : https://medium.com/@imanshumandubey/utilizing-deep-q-learning-to-make-a-chromes-dino-game-bot-d2c1e3d0308e