AI_for_Wechat_tiaoyitiao

AI_for_Wechat_tiaoyitiao

Artificial intelligence assistant automatically jumps to WeChat mini game

Author: Liu Yunfei

wechat:lyffly

0x00

Support mobile phones: Android Android system resolution is 1920 * 1080 pix
Computer: windows
What is the difference in the artificial intelligence version: I hope that the AI ​​understands the concept of distance. When jumping, it only jumps the approximate distance, not the center of each jump, not an exact value. This can also prevent points from being banned.

0x01

First configure the adb and Python environment as follows

https://github.com/wangshub/wechat_jump_game

0x02

Install tensorflow and keras

pip install tensorflow
pip install keras

0x03

Download all files here

0x04

Open the Jump Jump game in the downloaded folder and run it in cmd

python predict.py

Now you can start the game

See the video for the effect http://www.bilibili.com/video/av17998366

screenshot

If you just want to increase your score, don’t read any further.

0x05

Note: Except for the deep learning part, the code in this article is written by myself. Other codes come from https://github.com/wangshub/wechat_jump_game

The images used for training have been shared:
Link: https://pan.baidu.com/s/1mhCz1Bq Password: lih7

0x06

Knowledge background: deep learning, CNN
There are a total of 3502 training pictures, the accuracy is about 95%, the training graphics card (1050ti) takes about 10 hours.

The network structure is as follows

Continuously updating