During the New Year's holiday, playing cards and nagging is a must-have program. Even if you don't know much about playing cards, friends and relatives will push you to the table. You who work overtime 996 all year round and neglect your card skills will definitely become a rich boy as soon as you play. Directly refusing to go to the card table seems too petty, and it is hard to bite the bullet, but the wallet is also very specific. Is there any way to quickly improve your card skills and directly transform into a "gambling god"?
The answer may lie in the open source projects I recommend below.
The first, RLCard.
RLCard is a toolkit for the card game Reinforcement Learning (RL). It supports multiple card game environments including Landlord, Mahjong, Bridge, Texas Hold'em, and has an easy-to-use interface for implementing various reinforcement learning and search algorithms.
Game models supported by RLCard
You can use this open-source library to combine deep Monte Carlo, deep Q learning, virtual self-matching, and virtual regret minimization algorithms to play mahjong and landlords. Imagine if there were such a powerful algorithm to assist you in playing mahjong and fighting landlords, would you still be timid? I'm afraid you won't have enough hands then!
The AI Dou Di Landlord model in this project has an online experience address, if you want to experience it, you can leave a message to get it, and if you are interested, you can try it and feel the strength of AI Dou Di Lord.
The AI Doudi Master model has an online experience
At the same time, the children's boots of the Happy Fight Landlord can also pay attention to DouZero_For_HappyDouDiZhu project, using the DouZero model in the RLCard project above for the actual battle of the Happy Fighting Landlord, automatically identifying the hands and cards in the interface of the Happy Fighting Landlord program, and automatically playing cards according to the DouZero model prediction suggestions. Usually play Happy Fight Landlord, but pay attention, it is possible that your opponent is this AI!
RLCard is used to automate the Joy Fight Landlord
The second, playingcard.
Some students want to ask, the above RLCard is just an AI ah, can only provide suggestions, at most plus screen text recognition can be used to fight the happy fight landlord. But what about the real hand?
This is a card recognition project based on the yolov5 target recognition model, and the training data is basically the situation when we usually shake hands and cards at the table.
Yolov5 hand identification
Those with code capabilities can use this open source project and the above RLCard to combine it with a camera to take pictures. Real-time play and hand situation of the card table let playingcard perform target detection and recognition, and then input into the corresponding game AI model of RLCard, after getting the prediction suggestions of the AI model, tell you how to play cards through wireless headphones, reproduce the scene of AI guiding agents to kill all sides in "Suspect Tracking", so that you quickly change from a little white to a "gambler god".
Yolov5 object detection model for poker recognition
The third, Mortal.
This is an AI of Japanese mahjong. Why do I already have a mahjong AI model in RLCard, and I still want to bring the mahjong AI that has a good life here? Because for the complex incomplete information game of mahjong, it is still very difficult for AI to win humans steadily. But that has changed recently. Chaofeng AI from Microsoft has become a 10-dan player on the internationally renowned professional mahjong platform "Tianfeng", which requires Chaofeng to maintain a very high winning rate in tens of thousands of games. At present, there are about 330,000 players on the Tianfeng platform, of which only 180 have reached 10 stages. That is, 10 dan represents the top 0.0054% of the world's mahjong players,
At present, the highest ranking of mahjong AI is of course Microsoft's "Super Phoenix", and the second place is Japan's "Explosive Fight", of course, these top AIs are not open source. And this Mortal is an open source project that benchmarks "explosive". With AI to firmly control the hand, it is possible.
Open source mahjong model
However, the author still has to say that playing cards in the New Year can enhance the relationship between relatives and friends, but don't indulge.