datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
diecamera-crops
dieCamera — per-die crops
One cropped image per physical die, labelled with its type and face value. This is
the deliberately-simple training set for dieCamera's offline value reader — the app that
watches a dice tray and posts the roll into a virtual tabletop
(source).
For the full frames these crops were cut from (and the multi-die detector-training data), see
the companion repo G-G-Games/diecamera-frames.
Schema
Standard 🤗 imagefolder layout —… See the full description on the dataset page: https://huggingface.co/datasets/G-G-Games/diecamera-crops.diecamera-frames
dieCamera — physical dice, read by webcam
351 webcam frames of physical polyhedral dice on a tray, with per-die type, face
value and bounding box. Collected to train the offline reader in
dieCamera, an app that watches your dice tray
and posts the roll into a virtual tabletop.
1,079 labelled dice across the six standard types (d4, d6, d8, d10, d12, d20).
587 of those carry face values that are ground truth — confirmed by a human in the app,
or placed deliberately to a prompt.… See the full description on the dataset page: https://huggingface.co/datasets/G-G-Games/diecamera-frames.