[{"_1":2,"_155":-5,"_156":-5},"loaderData",{"_3":4,"_13":14,"_21":22,"_134":135},"root",{"_5":-5,"_6":-7,"_7":8,"_9":-7,"_10":11,"_12":-7},"currentUser","primaryEmail","siteVerification","6LeIxAcTAAAAAJcZVRqyHh71UMIEGNQ_MXjiZKhI","toast","enableDev",false,"incidentStatus","routes/_layout",{"_15":-5,"_16":17,"_18":17,"_19":11,"_20":11},"username","isAskGordonEnabled",true,"isUsingNewTemplate","showMinimalLanding","showHoneypotTrap","routes/_layout.r.$namespace.$reponame",{"_23":24,"_25":-5,"_26":11,"_27":11,"_28":29,"_30":31,"_32":-5,"_33":11,"_34":35,"_84":-5,"_85":86,"_125":-5,"_126":11,"_127":17,"_128":11,"_129":-5,"_130":11,"_131":-5,"_132":133},"canonicalUrl","https://hub.docker.com/r/hichtala/draw/tags","mcpServer","canSeeDockerfile","canSeeAutobuilds","namespace","hichtala","reponame","draw","cloudBuildSource","isAutobuildsAvailable","repo",{"_36":37,"_38":39,"_41":42,"_43":44,"_45":11,"_46":47,"_48":49,"_50":51,"_53":31,"_28":29,"_54":55,"_56":40,"_57":58,"_59":60,"_61":62,"_63":-5,"_64":58,"_65":66,"_67":11,"_68":29,"_69":70,"_75":11,"_76":77,"_81":29,"_82":83},"categories",[],"content_types",[40],"image","date_registered","2024-02-02T14:39:14.879735Z","description","DRAW - Detecting and Recognizing A Wild range of cards /\nOD model trained to detect Yu-Gi-Oh! cards","is_private","last_updated","2024-02-09T08:39:59.714669Z","last_modified","2024-10-16T13:48:34.145251Z","media_types",[52],"application/vnd.docker.container.image.v1+json","name","pull_count",2283,"repository_type","star_count",0,"status",1,"status_description","active","source","collaborator_count","full_description","

\n \n \n

\n\n\n[![Licence](https://img.shields.io/github/license/Ileriayo/markdown-badges?style=flat)](LICENSE)\n[![Docker Pulls](https://img.shields.io/docker/pulls/hichtala/draw?logo=docker)](https://hub.docker.com/r/hichtala/draw/)\n[![Twitter](https://badgen.net/badge/icon/twitter?icon=twitter&label)](https://twitter.com/tiazden)\n\n[![Model on HF](https://huggingface.co/datasets/huggingface/badges/resolve/main/model-on-hf-sm.svg)](https://huggingface.co/HichTala/draw)\n[![Dataset on HF](https://huggingface.co/datasets/huggingface/badges/resolve/main/dataset-on-hf-sm.svg)](https://huggingface.co/datasets/HichTala/yugioh_dataset)\n\n[🇫🇷 Français](README_fr.md)\n\nDRAW (which stands for **D**etect and **R**ecognize **A** **W**ild range of cards) is the very first object detector\ntrained to detect _Yu-Gi-Oh!_ cards in all types of images, and in particular in dueling images.\n\nOther works exist (see [Related Works](#div-aligncenterrelated-worksdiv)) but none is capable of recognizing cards during a duel.\n\nDRAW is entirely open source and all contributions are welcome.\n\n

\n\n

\n\nBoth a docker installation and a more conventional installation are available. If you're not very familiar with all the code, \ndocker installation is recommended. Otherwise, opt for the classic installation.\n\n#### Docker installation\n\nIf you are familiar with docker, the docker image is available [here](https://hub.docker.com/r/hichtala/draw).\n\nOtherwise, I recommend you to download [DockerDesktop](https://www.docker.com/products/docker-desktop/) if you are on Windows.\nIf you are on Linux, you can refer to the documentation [here](https://docs.docker.com/engine/install/).\n\nOnce it is done, you simply have to execute the following command,\n```shell\ndocker run -p 5000:5000 --name draw hichtala/draw:latest\n```\nYour installation is now completed. You can press `Ctrl+C` and continue to Usage section.\n\n\n#### Classic installation\n\nYou need python to be installed. Python installation isn't going to be detailed here, you can refer to the [documentation](https://www.python.org/).\n\nWe first need to install pytorch. It is recommended to use a package manager such as [miniconda](https://docs.conda.io/projects/miniconda/en/latest/). \nPlease refer to the [documentation](https://docs.conda.io/projects/miniconda/en/latest/).\n\nWhen everything is set up you can run the following command to install pytorch:\n```shell\npython -m pip install torch torchvision\n```\nIf you want to use you gpus to make everything run faster, please refer the [documentation](https://pytorch.org/get-started/locally/)\n\nThen you just have to clone the repo and install `requirements`:\n```shell\ngit clone https://github.com/HichTala/draw\ncd draw\npython -m pip install -r requirements.txt\n```\n\nYour installation is now completed.\n\n

\n\n

Usage

\n\nNow to use it you need to download the models and the data, in section [Models and Data](#div-aligncentermodels-and-datadiv).\n\nOnce you have it, follow instruction depending on you have docker or classic installation.\nPut all the model in the same folder, and keep the dataset as it is\n\n#### Docker installation\n\nYou have to copy the data and models in the container. Execute the following command:\n\n```shell\ndocker cp path/to/dataset/club_yugioh_dataset draw:/data\ndocker cp path/to/model/folder draw:/models\n```\n\nOnce it is done you just have to run the command:\n```shell\ndocker start draw\n```\nopen the adress `localhost:5000`, and enjoy the maximum. Refer [bellow](#both) for details about parameters\n\n\n#### Classic installation\n\nYou need to modify the `config.json` file by putting the paths of you dataset folder in `\"data_path\"` parameter \nand the path to model folder in `\"trained_models\"` parameter.\n\nOnce done, just run:\n```shell\nflask --app app.py run\n```\nopen the adress `localhost:5000`, and enjoy the maximum. Refer [bellow](#both) for details about parameters\n\n#### Both\n\n* In the first parameter, the one with gears, put the `config.json` file\n* In the second parameter, the one with a camera, put the video you want to process (leave it empty to use your webcam)\n* In the last one, put your deck list in the format `ydk`\n\nThen you can press the button and start the process !\n\n

\n\nIn this project, the tasks were divided so that one model would locate the card and another model would classify them. \nSimilarly, to classify the cards, I divided the task so that there is one model for each type of card,\nand the model to be used was determined by the color of the card.\n\nModels can be downloaded in Hugging Face. \nModels starting with `beit` stands for classification and the one starting with `yolo` for localization.\n\n[![Model on HF](https://huggingface.co/datasets/huggingface/badges/resolve/main/model-on-hf-sm.svg)](https://huggingface.co/HichTala/draw)\n\nFor now only models for \"retro\" gameplay are available but the ones for classic format play will be added soon.\nI considered \"retro\" format all cards before the first _syncro_ set, so all the cards edited until Light of Destruction set (LODT - 05/13/2008) set and all speed duel cards. \n\n

\n\n

Data

\n\nTo create a dataset, the YGOPRODeck api was used. Two datasets were thus created, \none for \"retro\" play and the other for classic format play. Just as there is a model for each type of card,\nthere is a dataset for each type of card.\n\nDataset can be downloaded in Hugging Face.\n\n[![Dataset on HF](https://huggingface.co/datasets/huggingface/badges/resolve/main/dataset-on-hf-sm.svg)](https://huggingface.co/datasets/HichTala/yugioh_dataset)\n\nFor now only \"retro\" dataset is available, but the one for classic format play will be added soon.\n\n\n

\n\n\n---\n##

💡Inspiration

\n\nThis project is inspired by content creator [SuperZouloux](https://www.youtube.com/watch?v=64-LfbggqKI)'s idea of a hologram bringing _Yu-Gi-Oh!_ cards to life. \nHis project uses chips inserted under the sleeves of each card, \nwhich are read by the play mat, enabling the cards to be recognized.\n\nInserting the chips into the sleeves is not only laborious, but also poses another problem: \nface-down cards are read in the same way as face-up ones. \nSo an automatic detector is a really suitable solution.\n\nAlthough this project was discouraged by _KONAMI_ ^®, the game's publisher (which is quite understandable),\nwe can nevertheless imagine such a system being used to display the cards played during a live duel, \nto allow spectators to read the cards.\n\n---\n##

🔗Related Works

\n\nAlthough to my knowledge `draw` is the first detector capable of locating and detecting _Yu-Gi-Oh!_ cards in a dueling environment, \nother works exist and were a source of inspiration for this project. It's worth mentioning them here.\n\n[Yu-Gi-Oh! NEURON](https://www.konami.com/games/eu/fr/products/yugioh_neuron/) is an official application developed by _KONAMI_ ^®.\nIt's packed with features, including cards recognition. The application is capable of recognizing a total of 20 cards at a time, which is very decent. \nThe drawback is that the cards must be of good quality to be recognized, which is not necessarily the case in a duel context. \nWhat's more, it can't be integrated, so the only way to use it is to use the application.\n\n[yugioh one shot learning](https://github.com/vanstorm9/yugioh-one-shot-learning) made by `vanstorm9` is a \nYu-Gi-Oh! cards classification program that allow you to recognize cards. It uses siamese network to train its classification\nmodel. It gives very impressive results on images with a good quality but not that good on low quality images, and it \ncan't localize cards.\n\n[Yolov8](https://github.com/ultralytics/ultralytics) is the last version of the very famous `yolo` family of object detector models.\nI think it doesn't need to be presented today, it represents state-of-the-art real time object detection model.\n\n[BEiT](https://arxiv.org/pdf/2106.08254.pdf) is a pre-trained model for image classification. It uses image transofrmers \nwhich are based on attention mechanism. It suits our problem because authors also propose a pre-trained model in `Imagenet-22K`.\nIt is a dataset with 22k classes (more than most classifiers) which is interesting for our case since there is mode than 11k cards in _Yu-Gi-Oh!_. \n\n---\n##

🔍Method Overview

\n\nA medium blog will soon be written and published, explaining the main process from data collection to final prediction.\nIf you have any questions, don't hesitate to open an issue.\n\n---\n##

💬Contact

\n\nYou can reach me on Twitter [@tiazden](https://twitter.com/tiazden) or by email at [hich.tala.phd@gmail.com](mailto:hich.tala.phd@gmail.com).\n\n","has_starred","hub_user","immutable_tags_settings",{"_71":11,"_72":73},"enabled","rules",[74],".*","is_automated","permissions",{"_78":17,"_79":11,"_80":11},"read","write","admin","user","storage_size",8056495086,"orgData","recentTags",[87],{"_88":40,"_89":90,"_91":92,"_93":94,"_95":96,"_97":98,"_46":114,"_115":90,"_116":29,"_117":52,"_53":118,"_119":120,"_121":105,"_122":114,"_123":62,"_124":17},"content_type","creator",25001635,"digest","sha256:7fcef5242cb5a78fc0e5fbe8754041fbfa5c62078c47aa548b7d7f9a12e2357e","full_size",4712989837,"id",596243757,"images",[99],{"_100":101,"_91":92,"_102":103,"_104":105,"_106":107,"_108":109,"_110":103,"_111":-5,"_112":94,"_59":62,"_113":-5},"architecture","amd64","features","","last_pulled","2026-09-24T02:16:07.783511363Z","last_pushed","2024-02-09T08:39:59Z","os","linux","os_features","os_version","size","variant","2024-02-09T08:39:59.029836Z","last_updater","last_updater_username","media_type","latest","repository",23099100,"tag_last_pulled","tag_last_pushed","tag_status","v2","sbxKitMetadata","isPotentialSpam","isHubSbxKitsEnabled","showSbxLogoOverride","sbxTrademarkText","displaySecurityMetadata","tagAttestations","tagsHaveAttestations",[],"routes/_layout.r.$namespace.$reponame.tags",{"_34":136,"_143":144,"_132":153,"_154":11},{"_36":137,"_38":138,"_41":42,"_43":44,"_45":11,"_46":47,"_48":49,"_50":139,"_53":31,"_28":29,"_54":55,"_56":40,"_57":58,"_59":60,"_61":62,"_63":-5,"_64":58,"_65":66,"_67":11,"_68":29,"_69":140,"_75":11,"_76":142,"_81":29,"_82":83},[],[40],[52],{"_71":11,"_72":141},[74],{"_78":17,"_79":11,"_80":11},"tags",{"_145":60,"_146":-5,"_147":-5,"_148":149},"count","next","previous","results",[150],{"_88":40,"_89":90,"_91":92,"_93":94,"_95":96,"_97":151,"_46":114,"_115":90,"_116":29,"_117":52,"_53":118,"_119":120,"_121":105,"_122":114,"_123":62,"_124":17},[152],{"_100":101,"_91":92,"_102":103,"_104":105,"_106":107,"_108":109,"_110":103,"_111":-5,"_112":94,"_59":62,"_113":-5},[],"canDeleteTags","actionData","errors"]

hichtala/draw

Tags for hichtala/draw