We implement a DQN (Deep Q-Network) agent to solve Unity's Banana Collector environment. The environment is solved in a small number of episodes with a remarkable high score and a stable learning process. We use Python as a language and Pytorch as a tool.