Mandine's Real Poker Hands (MRPH) Dataset

The MRPH dataset contains 2,277 real-world images of poker hands. Each image shows a complete 5-card poker hand photographed under realistic conditions. The dataset was created for computer vision and machine learning tasks such as poker hand classification. Dataset sources The images were collected from multiple sources: Poker Combos (Roboflow Universe) — 140 images Poker Cards (Roboflow Universe) — 176 images(only images containing exactly 5 cards were kept) Poker Hands by Doni Kurniawan (Roboflow Universe) — 507 images(only images containing exactly 5 cards were kept) Original photographs captured by the authors — 1,454 images Annotations Each image is annotated with a single label corresponding to the poker hand represented in the image. The dataset includes the following classes: RoyalFlush StraightFlush FourOfAKind FullHouse Flush Straight ThreeOfAKind TwoPairs Pair HighCard Notes All images contain exactly five playing cards. Images may contain some variations in lighting, background, card orientation, camera angle.