Today's article will show you the top 6 gaming datasets from Roboflow Universe to help provide inspiration for using video games or real world games with computer vision applications.
Using computer vision with gaming can lead to entirely new experiences:
- Gaming automation: using computer vision to automate actions in games.
- Mixed reality gaming: augmenting real world games to have digital components.
- Building game bots: using programming as a way to play a game.
Combining gaming and computer vision opens up new ways to interact with games and expands the gaming ecosystem with entirely new tools and mobile applications. Below are example datasets you could use to build a computer vision project for gaming.
1. Double Twelve Dominoes Computer Vision Project
Link: https://universe.roboflow.com/pip-tracker/double-twelve-dominoes
Project Type: Object Detection
Subject: Domino-pip-clusters
Classes: pip-1, pip-10, pip-11, pip-12, pip-2, pip-3, pip-4, pip-5, pip-6, pip-7, pip-8, pip-9
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The double twelve dominoes project is a dataset of 306 trains, 29 valid, and 15 test images of dominoes which can be used to maintain track of the final score after each round of gameplay. The dataset is resized to 416*416 pixels for better processing and has auto orientation applied. The images with different colour dots on dominoes are included in the dataset to detect different types and models of the game.
Test the model's performance by calling Roboflow's API pretrained on the images.
2. CSGO Train YOLO V5 Computer Vision Project
Link: https://universe.roboflow.com/new-workspace-rp0z0/csgo-train-yolo-v5
Project Type: Object Detection
Subject: players
Classes: CT, T, person
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
CSGO computer vision project contains 5.9K train, 431 valid and 37 test images of different players from counter-terrorist and terrorist teams from the Counter-Strike Global Offensive video game. The dataset has two classes for identifying the player type and each image in the dataset has data annotations applied. This dataset can be used for training bots to learn how to play the game. It will also be helpful for tracking, analyzing, and visualizing data for CSGO matches and competitions across multiple gaming platforms.
Test the model's performance by calling Roboflow's API pretrained on the images.
3. Draughts Board Computer Vision Project
Link: https://universe.roboflow.com/harry-field-qemqy/draughts-board-fm9sx
Project Type: Object Detection
Subject: Draughts-pieces
Classes: b, bk, bl, br, tl, tr, w, wk (
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
Draughts board computer vision project contains labeled images for capturing the game state on an 8x8 board. It can be used to develop a mobile droughts application enabling users to interact with their mobile device's camera. The dataset can also detect corner squares, and their board locations can be estimated. There are multiple classes captured by this dataset, including:
- White Pieces
- White Kings
- Black Pieces
- Black Kings
- Bottom left corner square
- Top left corner square
- Top right corner square
- Bottom right corner square
Test the model's performance by calling Roboflow's API pretrained on the images.
4. Chess Pieces Computer Vision Project
Link: https://universe.roboflow.com/joseph-nelson/chess-pieces-new
Project Type: Object Detection
Subject: pieces
Classes: bishop, black-bishop, black-king, black-knight, black-pawn, black-queen, black-rook, white-bishop, white-king, white-knight, white-pawn, white-queen, white-rook
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The chess pieces computer vision project is a dataset of 606 train, 58 valid, and 29 test chess board images and their various pieces. The images for building the dataset are collected with the help of a tripod at a constant angle and are annotated as:
- white-king
- white-queen
- white-bishop
- white-knight
- white-rook
- white-pawn
- black-king
- black-queen
- black-bishop
- black-knight
- black-rook
- black-pawn
The dataset can be used to build a custom object detection model to determine the state of the board at the beginning of a game and at any other instance during a competition or playing a general game. The project can also be extended to training bots to play chess fully autonomously.
Test the model's performance by calling Roboflow's API pretrained on the images.
5. Billiard_POOL Computer Vision Project
Link: https://universe.roboflow.com/kaji-hiroaki/billiard_pool
Project Type: Object Detection
Subject: Billiard-ball-annotation
Classes: b1, b15, b2, b4, b6, b7, b8, b9, cb
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The billiard pool computer vision project is a dataset of 3.1 trains and 786 valid images of different colored pool balls with the white cue ball. The images are resized to 416x416 pixels in dimensions, each with auto orientation applied. The dataset can be utilized to train a machine vision model capable of recommending a possible shot that can drop the desired ball into a pocket or automate the scoring of a game of pool.
Test the model's performance by calling Roboflow's API pretrained on the images.
6. UNO Cards Computer Vision Project
Link: https://universe.roboflow.com/joseph-nelson/uno-cards
Project Type: Object Detection
Subject: Card-Types
Classes: 0, 1, 10, 11, 12, 13, 14, 2, 3, 4, 5, 6, 7, 8, 9
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The UNO cards detection model contains 31K train, 1.8K valid, and 899 test images of UNO cards, each labeled and taken with different textual backgrounds. The dataset has many applications, from building an application for automatically counting scores for UNO matches to training a reinforcement learning agent to learn to play and master the game of UNO.
Test the model's performance by calling Roboflow's API pretrained on the images.
Using Open Source Gaming Datasets for Computer Vision
Computer vision will enhance the responsive and adaptive experiences for real world and virtual gaming applications, and we hope today's article provided you with useful datasets and ideas to apply new techniques to add new dimensions to the gaming experience.
To get started building a gaming computer vision project, sign up for a free Roboflow account to add images from open-source datasets and train a computer vision model in one click. You'll have a fully running computer vision pipeline in no time.