SoS is a semantic segmentation dataset for the RoboCup Humanoid League. It is semi-synthetically generated using Blender and real HDRs from past RoboCup fields.
There are 47,575 samples in the dataset. The dataset is split into ten partitions for easier download. A sample in the dataset is defined by the tuple containing the raw image, segmentation mask and metadata. Each folder contains full samples, that is the raw image, mask and metadata for one sample… See the full description on the dataset page:
https://huggingface.co/datasets/NUbots/SoccerSegmentation.
